When Integrated Services Digital Network is under discussion it’s generally the same topics that crop up – standards – or lack of them, when they’re coming, what’s wrong with them, and so on and so forth. What doesn’t usually get much of a mention are ISDN semiconductors. Basically, standards are thought to be controversial and therefore interesting, while chips – unless they are of the dumped variety – are not. It’s a bit like the foundations of a house – not generally as interesting to potential buyers as whether or not there is central heating, or en suite bathrooms – but pretty fundamental all the same. And judging by the claims of the major players, the ISDN semiconductor arena is not without controversies of its own. In value terms, too, the ISDN semiconductor market is certainly not to be sneezed at. Although it is expected to be worth just $17m this year, the growth rates are going to be phenomenal. US market research outfit In-Stat Inc reckons that the market is going to grow from its current level of $134m in 1989; $306m in 1990; $500m in 1991; 660m in 1992; 1,200m in 1993; $1,950m in 1994 and finally reaching a whopping $3,600m by 1995. Optimistic These correlate with calculations carried out by National Semiconductor Corp and SGS-Thomson SA. According to their research, there are around 25,000 ISDN hook-ups worldwide mainly in trials and mostly in the US and Europe. But as the pace hots up over the next four years, the number of lines shipped is expected to climb to around 5m. NatSemi reckons that in the early stages of ISDN each line will require around $140 worth of chips – making for a total semiconductor market of $700m by 1992 – pretty close to the In Stat $660 figure. Not everyone is so optimistic, however. Charles Louisson, European Strategic Marketing Manager, said that if these estimates were to be reached then 75% of all the line shipments in 1996 would have to be ISDN, … and clearly that won’t be the case. Not surprisingly, therefore, the world’s chipmakers have turned their attentions on this lucrative marketplace – mainly through partnerships. National Semiconductor and SGS-Thomson, for instance, already had a partnership that went back to 1979, which was extended in 1986 to cover ISDN, and is currently listing Northern Telecom Ltd and GEC Plessey Telecommunications among its customers. Meanwhile, Advanced Micro Devices Corp has a second sourcing deal with Siemens AG announced earlier this year, which is expected to deliver jointly developed products in due course – this partnership counts British Telecommunications Plc as a customer, and expects to announce two further deals within the next month. A third liaison is Mitel Corp’s deal with Level One Corp, under which the two agreed back in July to develop and second source advanced transceivers for T-1 transmission and ISDN markets. The NatSemi/SGS-Thomson camp sees the ISDN chip market evolving in stages, with the first large market for circuits most likely to be the main telephone exchanges of phone companies, which are moving into digital switching. In a main exchange, every line card that connects to a customer’s ISDN loop is a candidate for ISDN and the magic chips. Trung Nguyen, product marketing engineer at Telecom Products, listed the devices needed, starting at the main exchange, where a circuit called a SLIC – Subscriber Loop Integrated Circuit would be installed. Its function would be to provide the power. Continuing down the line comes what NatSemi call a Combo, which combines a speech digitising Codec coder-decoder chip with a filter. The filter cuts off the signal at 4KHz, making the sampling of the signal more manageable. A few weeks back the National Semiconductor – SGS – Thomson partnership launched the T3076/ST5075/6 Programmable Combo, a derivative of its second generation pulse code modulation Codec and filter devices. The fact that the Combo is programmable gives much greater flexibility, said Nguyen. For instance, users can specify the signal depending on how far away from the exchange it is and the degree to which it
would be degraded by the distance, or for times when a loudspeaker might be used, for example. He said that NatSemi/SGS-Thomson had been shipping low volumes of the device, which he expected to take over from the non programmable variety, and estimated that the European market for such a function is some 15m lines a year at $5 each. Intel, meanwhile, is reported to be shipping some 2,000 a week of its combo chip, the 29C48. Sometimes the voice digitising chip is combined with an S interface chip. AMD, for instance, has a product called the AM79C30A which does this, providing ISDN access for voice and voice+data terminal equipment. Other manufacturers, such as NatSemi, keep the S-interface separate. Its TP3420/ST5420A S device is a monolithic – single chip – transceiver that implements all the features needed for terminal equipment, terminal adaptor, NT1 and NT2 and PABX line-card applications. It addresses only Layer 1 of the ISDN standard stack, while some manufacturers’ products combine Layer 1 and 2 support in a single product. One such is the AMD device mentioned above – the AM79C30A. It meets the CCITT requirements for Layer 1 and some of the Layer 2, while other remaining Layer 2 and 3 requirements are implemented with a separate microprocessor. There appear to be two schools of thought on this. The NatSemi line from Trung Nguyen is that keeping the two separate makes the benefits of volume purchase are available to the user. Layer 2 support is not required for many applications such as network termination, so users generally need many more Layer 1 products than Layer 2. By combining both on a single product, users are being forced to buy redundant functionality, in other words complexity they do not need, which can work out more costly. Perspective AMD has a different perspective, expressed here by Charles Louisson: At the main switch or PABX level there are arguments for keeping them separate, but within a telephone it is much more cost-effective to keep the component count low. So while NatSemi would need three components within a phone – the S transceiver, the combo and the HDLC controller, we only need the one along with a separate microcontroller. The ISDN-specific elements are integrated so customers can take their choice of microprocessor. With the National solution they don’t have a choice – it is combined with the HDLC controller in their HPC16400 product. This is, however, set to change, with a new HDLC single-channel controller with Direct Memory Access due from NatSemi in first quarter 1989. This means it will be usable with standard 16-bit microprocessors – from any supplier.
