In the year 2000, the semiconductor industry has shown strong growth in all sectors, but one sector that has shown exceptional growth has attracted particular attention. That sector is flash memory - a type of electronic memory that is being increasingly used in a wide range of telecom, automotive, computer and consumer applications but which relatively few semiconductor manufacturers can produce in volume at the low cost that equipment manufacturers require.
Flash belongs to the 'nonvolatile' class of semiconductor memories - which retain the data that is stored in them while they are connected to power. Historically, volatile memories such as DRAM (dynamic random access memory) and SRAM (static random access memory) have been the most important types of memory in terms of market size, despite the disadvantage of their volatility, thanks to their very fast write capability which is mandatory for their use as 'working' memory. In the case of DRAM, this is because the small size of the memory cell allows very high storage capacities to be achieved. Today, 256 MbDRAMs are commercially available (250 million data bits per chip) and 1 Gb devices (1000 Mb) will soon be introduced. For SRAM, the advantage is less overhead (no refresh circuitry) and also, for some types, the speed at which data can be read from the memory; this is typically an order of magnitude faster than DRAM. Both DRAM and SRAM are essential components of PCs.
EPROM? EEPROM? Flash?
The first important nonvolatile memory technologies were EPROM and EEPROM. An EPROM (erasable programmable read-only memory) allows data to be written to the memory once and subsequently read out any number of times whereas an EEPROM (electrically erasable PROM) is more flexible in that it allows individual memory locations to be erased and reprogrammed many times but the penalty for this flexibility is a more complex memory cell that increases the cost and reduces the storage capacity per unit chip area. For this reason, EPROM is mainly used as an easy method for storing large amounts of reasonably stable program code and EEPROM for storing parameters and other information that needed regular updating.
Flash applications
Flash technology originally entered the market as an alternative to EPROM. Like EEPROM, a flash memory can be electrically erased and reprogrammed with virtually no practical limit on the number of erase/program cycles but, unlike EEPROM, flash memories can be made at low cost with very large storage capacities. Flash also has the additional advantage that, unlike EPROM, it is not necessary to erase the whole memory array in order to store new data in part of it; flash memories are typically organised as a number of sectors, each of which can be individually reprogrammed.
The first applications for flash memory were primarily as code storage, where the greater flexibility compared to EPROM outweighed the additional unit cost. More recently, flash has opened up new applications where its key function is to store large amounts of data such as the MP3 music files or digital image files that are increasingly produced in digital consumer applications.
Market info
In 1999, the market for flash memories grew by 83% to $4,6 bn, overtaking SRAM to become the second biggest memory market after DRAM. The flash market (including both standard products and special architectures) will continue this vigorous growth for several years; market researchers IC Insights predict that the market will more than double in 2000 to exceed $10 bn and that average selling prices (ASPs) will continue to rise until 2002.
Unlike DRAM, flash memories are used in a wide variety of applications such as cellular phones, hard disk drives, set-top boxes and engine management units. The key applications that are driving the market growth are portable applications. This is because flash memories offer a combination of nonvolatility, high storage density, low cost and low power consumption that no other memory technology can match today. The amount of flash memory used in applications is rising; for example, mobile phones typically required 4 Mb in 1998 but this could grow to as much as 64 Mb in 2002.
Today, the flash market is limited not by demand but by capacity. Flash memories are significantly harder to manufacture than DRAM and there are relatively few chip manufacturers with the know-how required to produce flash memories in volume.
Flash also differs from DRAM in that the market has evolved into three different product groups, each with its particular cost/performance optimisation: commodity stand-alone flash memories, embedded flash and, most recently, application-specific/advanced architecture devices. For standalone and application-specific products, 0,25 µm was the state-of-the-art production process earlier this year but this will soon move to 0,18 µm. For embedded flash products, 0,35 µm processes are in volume production today, with 0,18 µm expected in the first half of 2001.
The flash challenge
The pressures on flash manufacturers are twofold: to increase performance (in terms of speed, density and power consumption) and to decrease costs. ST uses two complementary approaches to achieve these goals. The first is the continual scaling down of process geometries, not only via the technology nodes defined by the semiconductor industry's ITRS (International Technology Roadmap for Semiconductors) roadmap (0,25 µ, 0,18 µ, 0,13 µ, 0,10 µ ...) but also by process shrinks to intervening geometries such as 0,15 µ. The second is the development of new memory architectures that allow the current volume production technology to be optimally exploited in specific applications.
Because the move from one technology node to the next involves a concerted worldwide effort by semiconductor manufacturers and their suppliers of equipment, materials and CAD tools, individual chip manufacturers cannot advance independently and leapfrog the competition. Instead, the whole industry progresses with an interval that has moved in recent years from three years to two years between technology nodes. The challenge for individual chip manufacturers lies in mastering each new technology node, bringing it from R&D to profitable volume production as quickly as possible.
The ITRS roadmap for memories is based on DRAM, as this is the largest memory market and critically dependent on minimising device geometries. Because flash is harder to produce than DRAM, the ITRS roadmap assumes that flash memory technology will lag at least one generation behind DRAM eg when DRAM is produced in 0,18 µ, flash will be produced in 0,25 µ.
ST has pursued a more ambitious goal, introducing three new flash technologies in 18 months, with the aim of aligning its flash roadmap with the ITRS DRAM roadmap. For example, ST's investment in the recently opened R2 Technology Centre in Italy reflects ST's belief that flash can match DRAM in terms of design rules, density and cost. Flash memory size will surge to 1 Gb in the next few years, matching DRAM densities through the use of techniques such as multibit/cell.
Differentiated flash memories
The second approach that ST is successfully pursuing - developing application- specific architectures that allow current flash production technologies to be optimally exploited - depends on close cooperation with the major OEMs that are driving the development of new markets. This is an area where ST has a considerable advantage over other flash suppliers because it has strategic alliances with key players in the computer peripheral, communications, automotive and consumer segments.
The mobile phone market illustrates the explosive growth expected in the flash market. Initially, mobile phones contained fairly small flash memories (1-4 Mb) that were used for code storage. In the next five years, the functionality of the mobile phone will explode as third-generation (3G) phones introduce multimedia services such as Internet browsing and gaming, GPS, interactive news delivery, videoconferencing and CD-quality music. Each new service will increase the amount of flash memory needed for code storage, which will increase to as much as 128 Mb in this period. At the same time, equally large amounts of flash memory will be required in the phone for mass storage of data such as downloaded music or image files. This means that a single mobile phone in 2004 could contain as much as 128 MB of flash memory, compared to around 8 Mb today.
More importantly, standard flash memories will not be the best choice for these products. The mutually exclusive requirements of maximising performance and minimising both cost and power consumption can only be partly met by moving to new technology generations. Thus, the demands of the mobile phone market will be best met by flash memories that have been tailored to that particular application.
A similar growth can be foreseen in the set-top box and DVD market, with typical Flash content rising by an order of magnitude as STBs become more sophisticated. Again, to meet the demanding requirements of this market, which is both highly cost-sensitive and highly performance-sensitive, special flash memory architectures optimised for this application (eg x32 dual bank architectures with 100+MHz burst read performance) will be used rather than standard flash memories.
Co-operation
Working with leading OEMs has helped ST to develop innovative products such as ultra-fast (25 ns!) access time flash memories for hard disk drives and the world's first 32 Mb flash memory to combine a dual bank architecture and fast page mode access with full functionality with a 1,8 V power supply - ready for the next generation of cellular phones.
The leading position it is establishing in differentiated flash memories is allowing ST to add significant value to its products: by working with end users to develop application-specific flash architectures, ST is able to leverage its deep understanding of flash technology, its manufacturing capability and its system know-how to create products that give its customers a competitive edge.
For further information about STMicroelectronics memory products contact Arrow Altech Distribution (011) 923 9600, AGAtronics (011) 789 1065, Avnet Kopp (011) 444 2333, or Communica (012) 322 7613.
| Tel: | +27 11 923 9600 |
| Email: | [email protected] |
| www: | www.altronarrow.com |
| Articles: | More information and articles about Altron Arrow |
| Tel: | +27 12 657 3500 |
| Email: | [email protected] |
| www: | www.communica.co.za |
| Articles: | More information and articles about Communica |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version