PCMCIA

The New Interface Standard for Mobile Computers

The broad acceptance of the Personal Computer Memory Card Association (PCMCIA) standards for mobile computer peripherals has opened the floodgates to a wide range of new add-on products. Already, manufacturers are providing plug-in PC Cards containing memory expansion, mass storage and software for the growing array of notebook, sub-notebook and palmtop computers, as well as the newer personal digital assistants. Soon, hard disk drives, modems, LAN interfaces and other devices will appear, providing mobile computers with the power and flexibility previously available only on desktop machines.

The PCMCIA standards are new and still evolving, however, and implementing them presents many challenges to developers of PCMCIA components. PC cards must deliver their functionality in a very limited space while being able to withstand use in harsh environments. They must maintain acceptable performance while conserving battery power. And host systems must work with a wide range of peripherals, even in the absence of a precise definition of compatibility.

This paper takes a look at the PCMCIA standard's impact on the mobile computer market and the technical issues surrounding it.

The Mobile Computer Market

Since the introduction of the laptop computer, technology's constant drive for smaller, faster, cheaper products has given rise to a growing range of computers that are both smaller and more mobile than the laptop. Besides the notebook, sub-notebook and palmtop computers already on the market, a new generation of hand-held computers and personal digital assistants -- combining the functions of personal organizers, pagers and cellular telephones -- are starting to bring the advantages of mobile computing to an entirely new class of users, and represent the fastest growing segment of the computer industry.

The new mobile computers are designed to be used by those who are mobile while performing their work. These miniature devices will be used by sales people meeting with clients, doctors and nurses making hospital rounds, insurance claims adjusters at customer sites, truck drivers on the road and warehouse workers performing inventory control. Laptop and notebook computers, while portable, still function largely as desktop computers. They merely move the desktop to a hotel room or an airport lounge. Consequently, the new mobile computers will have to be smaller, lighter, more rugged, easier and less intrusive to use, and will frequently need equally portable and flexible access to remote databases and host computers through modems, wireless communications and LANs.

Mobile computers are available now, and the market is growing rapidly. According to Lempesis Research, laptop, notebook and pen-based computers currently make up approximately 22 percent of U.S. personal computer shipments, with desktop computers accounting for the rest. This figure is expected to increase to just over half of all PC shipments in three years. And pen-based computers, the most mobile portion of the overall portable market, will sell 5.4 million units in 1995, or roughly 40 percent of all portables shipped.

The Standard Approach

The mobile computer market is still so young that the market's needs are still not well defined. Neither users nor manufacturers know what specific combination of features will best suit how these machines will be used over the long run. As a result, manufacturers of mobile computers need a way to provide machines that are flexible and adaptable while also maintaining compatibility with the wide range of peripherals, add-ons and software that users might reasonably expect to incorporate into their systems.

For years, desktop systems have accommodated a broad range of peripheral devices through expansion slots, providing a standardized mechanism for users to add whatever devices they needed. These slots helped to create a huge market for add-ons, reducing their costs while increasing the variety and capability of devices available.

Until now, no similar expandibility option has existed in the mobile computer market. Existing desktop solutions, including the ISA, EISA, MicroChannel and SCSI busses, were not adaptable to the mobile computer form factor. An entirely new approach was needed.

To address this expandability problem, hardware and software developers reconsidered system architectures for the first time since the personal computer revolution began. They needed to devise a way to allow users to add functionality to their systems without having to open the case or know anything about the underlying technology. The method had to support a wide array of data storage and peripheral devices -- from memory cards and modems to hard disk drives, LAN interfaces and other functions -- with low power consumption, fast data access speed and compact size. And it had to be rugged enough to withstand operation outside the traditional office environment.

Unable to use the same expansion slots as their desktop cousins because of size constraints, and lacking any compelling standard, mobile computer manufacturers resorted to proprietary and incompatible solutions that severely segmented the market for add-on devices and limited their availability. Recognizing this growing impediment to market acceptance, manufacturers began to forge a single specification through the PCMCIA, seeking a standard approach that would allow more hardware and software to work together, lower costs for developers and users and reduce the risk of investing in obsolete or unsupported systems and components.

Originally developed in 1985 by the Japanese Electronic Industry Development Association (JEIDA) to provide a method for expanding memory in small form factor devices such as instruments and hand-held terminals, the PC Card standard has evolved through joint development of the standard by PCMCIA and JEIDA to include input/output functions common to personal computers.

The PCMCIA standard today specifies the physical dimensions, pin assignments, electrical characteristics, protocols and file formats for removable, credit-card-size devices. The standard provides hardware specifications for both PC Cards (memory and peripherals) and the hosts the PC Cards plug into. Taken together, this standard provides an extension of the native x86 CPU bus that allows PC Cards and system data to be interchanged -- a variety of peripherals can be inserted into any PCMCIA slot in any PCMCIA host, mobile and desktop, regardless of the underlying hardware or operating system.

Additionally, PCMCIA specifies a host software environment capable of accommodating a wide range of peripheral devices, including those yet to be developed. The result is a flexible, expandable peripheral standard that for the first time creates a link between I/O hardware and interface software.

Challenges of Implementing the Standard

The resulting PCMCIA 2.0 standard establishes a framework for developing both memory and I/O devices that can be plugged into and out of a system as easily as changing a cassette tape. Developing the technology to implement the standard, however, is not so easy, and provides a number of challenges for host system and peripheral developers. They, in turn, are looking to component vendors to provide solutions in the form of host bus adapters and peripheral controllers designed specifically to support the PCMCIA standard in a mobile environment.

Because the components supporting the PCMCIA standard will operate in a mobile environment, greater demands are placed on them than would be for equivalent components on the desktop. In addition, the mobile computer market is highly competitive, and system and peripheral developers will be seeking components that also provide them with a competitive advantage. As a result, component vendors need to address several technical issues related to both PCMCIA and mobile computing. The major issues are:

How these major issues are addressed will also have an impact on cost, compatibility and flexibility, as these issues are all interrelated. Power consumption, flexibility and performance are interrelated, for example, as are cost and form factor.

Power Consumption

One of the greatest limiting factors confronting mobile computer and peripheral designers is battery life. The less power a mobile computer consumes, the longer it can operate without replacing or recharging the batteries. As a result, mobile computers place stringent power constraints on their subsystem components to maximize battery life and minimize heat dissipation.

The two major ways to conserve power is power management -- shutting off components that are not being used -- and changing to 3.3-volt operation, which can reduce power consumption by 50 percent over traditional 5-volt operation. But the PCMCIA standard requires 5-volt operation, at least for initializing PC Cards, thus encouraging developers of host bus adapters and peripheral controllers to provide dual-voltage operation for maximum power conservation. Unfortunately, today's PCMCIA host bus adapters and peripheral controllers offer nothing in the way of power management or dual-voltage operation.

Careful design of PCMCIA components can enable them to conserve power through a range of operation-limiting power-down modes and dual-voltage operation. A host bus adapter or controller needs to offer system and peripheral designers the greatest flexibility in choosing ways to conserve power while taking into account the ease of entering and exiting power-down modes, as well as the level of intrusiveness it presents to the user. Through successive levels of turning various portions of the system off, power can be conserved while maintaining the level of operation the user desires.

When running a software application, for example, it is not uncommon for relatively long periods of time to pass when, from the perspective of the controller, nothing is happening. The application may be running in a tight loop waiting for keyboard input, or updating the display. During these times, the PCMCIA peripherals are not being accessed.

If the PCMCIA host adapter and peripheral controllers could automatically enter a state of limited operation during these periods, power would be conserved. During these times, the controllers' internal clocks can be turned off to most of the chip and the PCMCIA address and data lines can be set to a static value. Sometimes called standby mode, this power saving technique can be initiated automatically by the controller after a predetermined period of inactivity, and ended when access resumes, without CPU intervention.

Suspend mode gives the user a way to halt program execution in mid stream, placing the computer in a power-saving state, and resume full operation at a later time from the same point. Entering spreadsheet data may get interrupted by boarding a flight, only to be continued once seated and airborne.

In addition to the steps taken in standby mode, suspend mode virtually shuts down the controller, ignoring accesses made to any connected PCMCIA devices and stopping the controller's clock completely. An intelligent controller can handle the entire process when initiated by a hot-key combination, for example, or closing the mobile computer's cover.

A mobile computer placed in suspend mode may wind up remaining there for more than a short time. Left in this mode too long, the machine would only waste precious battery power, albeit at a much reduced rate. With the increased use of CMOS static memory and CPUs with fully static operation, mobile computers can be made to virtually turn off without losing valuable data or their place in an executing program. It is vitally important, therefore, that PCMCIA controllers support this capability by providing access to full information about their status prior to shutdown so this information can be saved and later restored after power up.

Flexibility

The movement to 3.3-volt operation will not occur overnight. Instead, systems will evolve from 5-volt to 3.3-volt operation, with many systems combining elements of both for some time to come. This has a direct impact on the PCMCIA controllers.

Hard disk drives, for example, operate primarily at 5 volts, as do many of the support chips around them. Yet other chips may be available in 3.3-volt versions. Eventually, 3.3-volt drives will appear, too. A manufacturer of a hard disk PC Card, therefore, will want the greatest flexibility in a hard disk drive controller, seeking one that can operate in any combination of 3.3- and 5-volt environments, to minimize design changes as lower power components are used.

The PCMCIA specification requires the ability to insert or remove PC Cards at any time, even while the computer is turned on. In order to allow for PCMCIA cards with unknown I/O functionality to be connected to appropriate non-conflicting interrupt locations requires a host bus adapter with flexible interrupt handling capability. The adapter must also be flexible in its approach to mapping memory and I/O locations to the system address space in order to maintain compatibility with existing PC hardware and software. For example, a modem card should be accessed by standard communications software as though it were at COM1 or a similar standard location. This necessitates locating the modem at I/O addresses from 2F8 to 2FF with interrupts generated on the IRQ4 line.

Form Factor

A single-chip controller does not mean a single-chip solution. Some "single-chip" controllers actually require external line buffers and other "glue" logic to integrate them into a system, and as a result require more board real estate. Only after considering the total number of chips needed for a complete solution, the types and sizes of the chips required and the efficiency of the layout can a fair assessment of form factor be made.

First, the total number of chips required must be considered, including any additional chips needed by the controller chip set. Some host adapters, for example, require CPU bus interface support logic and external line buffers to drive the PCMCIA slots.

It is also important to look at the type of packaging used for the controller chips. The requirements of the PCMCIA form factor push current packaging technology to its limits, raising the question of chip quality. Are the chips thin enough to be placed on the board without expensive cutouts, and if so, do they achieve the required mechanical integrity and environmental stability to be reliable over time? And can this be achieved at a reasonable cost?

Finally, after all the other considerations have been made, it is still necessary to explore how closely together all the chips involved can be placed. This is a function of the controller's pinout, where the various functions are located along the periphery of the chip. Unfortunately, pinout is usually done at the convenience of the chip designer without considering how the chips will be used. An intelligent approach to pinout, however, groups pins with related functions together, and offers efficient board layout, avoiding the need to route signals around the chip. Signals connecting to the slot interface would be grouped together with convenient access to the slot, while hard disk drive control signals, for example, would be placed elsewhere.

Performance

The power constraints placed upon PCMCIA controllers have a direct adverse affect on their performance. The dramatic power savings achieved by reducing operating voltage to 3.3 volts instead of 5 volts also reduces the speed at which a controller's clocks can operate, reducing performance. To maintain the desired performance, component manufacturers will need to turn to architectural changes.

Traditional 5-volt hard disk controllers, for example, are able to achieve adequate throughput using a single-bit read/write channel for data to flow between the controller and the hard disk drive. Operating at 3.3 volts, however, turns this single-bit channel into a bottleneck, requiring the use of multi-bit channels operating in parallel to compensate for the performance hit resulting from the reduced clock rate.

Performance at 3.3 volts also can be improved by reducing the dimensions of the chip itself. This is accomplished by reducing the size of the individual features of the chip, such as the size of the transistors and the thickness of the lines connecting them. But reducing feature size is not a straightforward process like reducing a document on a photocopier. Many component vendors do not retain complete control over how their design will be implemented in silicon, relegating that process to third-party fabrication facilities who create the photographic mask from the vendor's design data and then manufacture the chip. Those component vendors maintaining control of mask generation will have a clear advantage in providing reduced-feature products.

Architectural changes are also required to accommodate PCMCIA's requirement for access to setup data stored in memory on each PC Card. While it would seem to make sense for a hard disk drive controller implementation, for example, to place this data area in a convenient portion of the buffer already being used for disk data, this approach could result in access conflicts as both the drive controller and the PCMCIA host adapter contend for access to the buffer. Adding a dedicated memory area for the PCMCIA data on the hard disk controller eliminates this contention problem.

Host bus adapters can greatly improve system performance by providing features geared to the requirements of both the host system and the PCMCIA peripherals. A write buffer between the host CPU and slower PCMCIA peripherals could provide zero-wait-state performance by allowing the CPU to write to the buffer instead of directly to the peripheral. An intelligent controller could provide smart bus sizing so that 8-bit PC Card accesses can occur at faster 16-bit cycles on the host side.

Timing values required for setup, active and recovery times to access a PCMCIA card can vary considerably from card to card depending on the type of peripheral being accessed. Flash cards are different from hard disk drives and modems. Maximum performance is achieved when the host adapter can be programmed to match the timing requirements of each individual PCMCIA card, relieving the host of this chore. Combining this capability with the write buffer would enable flash cards, for example, to be programmed in background by eliminating the overhead of timing loops.

Putting it All Together

Creating host bus adapters and peripheral controllers for implementing PCMCIA systems is not a simple matter of making existing chips smaller. It is a complex task that requires the component designer to anticipate a number of unkowns and offer system and peripheral developers the greatest flexibility to deal with an evolving standard. Obviously, chip manufacturers who have experience creating leading-edge technology will be in the best position to contend with the uncertainties surrounding implementation of PCMCIA-compliant peripherals and systems.

Cirrus Logic understands these issues. Since it's beginning, Cirrus Logic's mission has been creating innovative peripheral controllers to meet the needs of transitional markets, and Cirrus Logic has the best track record in the industry. It is now poised to provide creative solutions for PCMCIA systems and peripherals.

# # #