Pick a card: card formats

The diversity of available memory cards has long complicated your product-definition and -design decisions (Reference 1). But, even though smaller and more obscure formats have faded over the years from the scene, even more formats have replaced them, and the contenders that survived have diversified into copious permutations (Figure 1 and Table 1).

Blame capitalism. At the beginning stages of the life cycle of an application, such as digital imaging, the free-market system encourages a diversity of incompatible approaches to problem-solving, with each alternative striving to get the biggest slice of the overall application pie. In the near term, the resultant confusion encourages consumers to keep their money in their wallets, constraining the size of the pie. Eventually, however, some of the problem-solvers run out of money or otherwise lose enthusiasm for the game. Others carve out a sustainable niche, often with the help of "bundling" partnerships with other companies and sibling divisions of the same company targeting the same application—digital cameras, for example. The market settles down to a few alternatives, and the pie grows. Then, a new application bursts onto the scene, and the race is on again.

Time and again, a consistent theme in the card-format industry market emerges. If several competing approaches exist, all able to meet the application needs, but one is more compact than the others, the industry will anoint it as the dominant choice. The corollary to this theorem is that, once a more svelte suitor can fully address application requirements, it'll capture the lion's share of the business that its predecessor previously owned. The driving force behind this trend is simple: Vendors are squeezing more and more functions into smaller and smaller system form factors. But the words "able to meet the application needs" are critical.

Is a given format dense enough and at a cost-effective price point? Does its system interface deliver adequate burst and sustained read- and write-transfer rates (see sidebar "Who's write?")? Is it sufficiently rugged and power-thrifty? If removable, is it so small or thin that a consumer could accidentally break it or leave it in a shirt pocket that subsequently goes through the laundry? Costs may be the first 10 priorities that an engineer uses as evaluation criteria, but they aren't the only priority. Consumers intuitively realize, if not right away than assuredly after their first debacle, that the value of the information on memory cards is much higher than the price they pay for the memory cards, thereby explaining the fast-growing market for image and other file-recovery utilities for memory cards. Therefore, as a rule, they're less price-sensitive in this arena than they are in other shopping situations.

The fading fortunes of the memory-card trailblazer, PCMCIA, or PC Card, aptly exemplify the form-factor-cross-over rule. I was on the team in the early 1990s that developed the ExCA (Exchangeable Card Architecture) standard at Intel, which eventually became industry-standard PCMCIA. With great fondness (and a little embarrassment), I remember how proud we all were of the first-generation Intel flash-memory cards with their then-whopping 1-Mbyte capacities. Now, of course, you can buy 8-Gbyte PC Cards based on flash memory from Pretec and other companies, and 16-Gbyte versions are scheduled to appear this year. But few consumers will buy these products because most applications don't demand these hefty densities, and users are unwilling to pay for them. Look at Toshiba. The company offers 1.8-in. hard-disk drives in densities as large as 60 Gbytes, yet the largest PC Card it produces based on those drives stores only 5 Gbytes.

PC Card slots are also rapidly disappearing from notebook PCs, as built-in connectivity broadens its penetration and as smaller, cheaper alternative expansion connectors become mainstream (see sidebar "I/O devolution"). For these and other reasons, CompactFlash cards, smaller derivations of the PCMCIA and follow-on CardBus hardware and software interfaces, have taken over most of the PC Card business, greatly expanding the overall card market in the process. Until recently, CompactFlash was the dominant card form factor. Yet, it too is in the process of becoming obsolete, as even smaller formats break through the good-enough barrier. 1-Gbyte CF cards based on flash memory are mainstream, larger versions are also available for niche applications, and 4-Gbyte, 1-in. hard-disk drives (with 5-Gbyte versions on the way from Seagate) squeeze into the Type II CF-format option. Cards based on magnetic media are traditionally cost-effective only at the highest densities in the available capacity range, especially when they come in nonstandard form factors (see sidebar "Surgical success").

Over time, the flash-memory-versus-hard-drive decision criteria remain constant, with flash getting the nod at low densities, thanks to its superior power-consumption and ruggedness characteristics. However, the one-versus-the-other crossover point in density continues to migrate upward. To the extent that hard-disk-drive densities overshoot an application's requirement, the semiconductor-versus-magnetic-storage pendulum shifts in flash memory's favor. Dense cards are popular with "prosumer" (professional/consumer) and professional-camera users, for example, who take lots of high-pixel-count, raw, and TIFF pictures. Sooner or later, though, flash-memory capacities will adequately serve even that need, elbowing aside the hard-drive alternative. And, smaller card form factors will grow in capacity to the point at which they will push aside CompactFlash. It's survival of the fittest.

As the CompactFlash market matures, suppliers are differentiating their products in the hope of eking out continued profits. Lexar Media, for example, has added WA (write-acceleration) technology support to its cards' memory controllers; if the mating system has WA-aware firmware that employs Lexar-friendly superset algorithms, Lexar claims, a meaningful performance boost will result. Sandisk's 4-Gbyte cards neatly circumvent the 2-Gbyte-capacity limitation of the antiquated FAT (file-allocation-table)-16 file-system support that many systems include. You can operate the card at its full single-partition capacity for use with FAT32. (Microsoft doesn't recommend NTFS-formatting flash-memory-based cards due to the file system's excessive "thrashing" of the media.) Alternatively, you can appropriately toggle a mechanical switch to access two 2-Gbyte partitions. Acknowledging the growing prevalence of single-slot CompactFlash systems, Sandisk also sells CF cards that combine flash-memory storage with IEEE 802.11b connectivity. And Lexar and Sandisk, along with other companies, offer cards rated for extended-temperature storage and operation.

Reference 1 predicted SmartMedia cards' eventual demise, and this forecast has largely come to pass. SmartMedia cards, formerly known as SSFDCs (solid-state floppy-disk cards), are in effect just other packaging options for the NAND flash-memory die inside them. They don't contain a memory controller, which their advocates claimed translated to lower media cost. Although this situation may have been true in the mid-'90s, Moore's Law-driven less-than-100-nm lithographies have made it a hypothetical point, and, overall, the disadvantages of the controllerless approach outweighed the advantages. Consumers could not upgrade a system-housed memory controller for compatibility with subsequent-generation memories that had higher densities, altered program- and erase-block sizes, and the like. Other flash-memory technologies could not easily shoehorn into the NAND-centric SmartMedia interface. And people commonly viewed SmartMedia as too flimsy to withstand the kind of abuse that they would subject it to.

The CompactFlash card's loss of momentum has been, to a large extent, the SD (formerly, Secure Digital) card's gain. SD's success story is an example of the classic adage of being in the right place at the right time, helped along by a heaping portion of politically fueled partisanship. Ironically, few of today's applications harness the "secure" aspect of SD—specifically, its support for CPRM (content protection for recordable media, Reference 2). However, the SD card, which Matsushita/Panasonic and Sandisk originally championed, burst on the scene right when the point-and-shoot-digital-camera and PDA markets started to heat up. Both applications welcomed its combination of thumbnail-sized form factor and higher-speed-than-the-alternatives, 4-bit data-transfer bus (see sidebar, "Testing, testing").

SD cards in densities up to 512 Mbytes are now in high-volume production, and 1-Gbyte cards have recently appeared. Like its CompactFlash predecessor, the SD card comes in both conventional and enhanced variants; the enhanced versions tout such features as ultrafast read and write performance and tolerance of exposure to water and extended temperature. Sandisk is, as with CompactFlash, bundling WiFi connectivity and flash-memory storage in a common SD card for single-slot systems containing SDIO controllers. One application that SD slots haven't widely penetrated, however, is cell phones—aside from PDA-derived Pocket PC Phone Edition units. Part of the reason is that, until recently, with the emergence of camera-inclusive phones, no compelling reason existed for large amounts of resident storage.

The other key aspect of the SD Card's lack of success in phones centers on the phone manufacturers' intolerance for paying royalty fees. Enter the knight in shining armor: the royalty-free MMC (MultiMediaCard; association membership includes license fees). MMC's lineage goes back to then-Siemens' (now Infineon's) ROM-based card of the mid-1990s. Ironically, because the nine-contact SD card is a functional superset of the seven-contact MMC, MMCs could find use in any SD application. They haven't to any significant degree because of several factors, most notably, speed: MMC's single-bit data bus, running at 20 MHz, allows for only a 2.5-Mbyte/sec peak transfer rate; SD's peak transfer rate is four times higher. Other factors include that promotion of the product has to date been relatively invisible and that industry-standards bodies work at a molasses-slow pace. How slow? My four-year-old article mentions an under-development specification enhancement, security, which the MMC Association still hasn't finalized.

The MMC Association claims, however, that it's now charging full steam ahead. It recently finalized Version 4 of the MultiMediaCard specification, also known as MMC Plus and HS (High Speed)-MMC, which evolves the bus into backward-compatible 1-, 4-, and 8-bit-wide options. (The 4-bit option is hardware-compatible but software-incompatible with the SD card implementation.) The 8-bit option translates to a 13-contact card interface and boosts the clock rate to 52 MHz, thereby delivering a peak transfer rate as high as 52 Mbytes/sec. Dual-voltage cards run in both 1.8 and 3V systems. And MMC Association Executive Director Andy Prophet predicts that the association will by year-end finalize the Secure MMC specification, which it is developing with the OMA (Open Mobile Alliance). Then again, the MMC Association's press-release archive reveals that the organization also claimed it had in late 2001 completed the secure-card specification. Renesas has also recently unveiled its proprietary X-Mobile Card, which adds full smart-card functions, such as a Java Card operating system, hardware tamper-resistant module, and cryptographic functions, to the MMC form factor.

One other roadblock in the way of SD Card or MMC adoption in diminutive cell phones is their proportional size. In response, both associations have developed smaller card variants, respectively called the mini-SD and RS-MMS and renamed the RS-MMS as MMC Mobile in its Version 4 high-speed interface version. A removable-memory card isn't the only architectural option cell-phone manufacturers are considering for boosting local-storage capacity. And it's not clear that the cellular carriers will be interested in offering phones with large amounts of resident storage. They may prefer that camera-phone users, for example, upload pictures to the carriers' servers for subsequent editing, e-mailing, and printing, thereby both increasing consumers' airtime usage and creating an additional service-revenue stream. Although the carriers' vision in theory sounds good, at least for the carriers' stockholders, the practical limitations of sluggish upload speeds may dash those plans.

At first glance, you might think it strange to see the words "limited adoption" referring to the small-form-factor Memory Stick format. After all, several memory-card manufacturers' portfolios promote it, and a diverse collection of cell phones, PDAs, still cameras, printers, videocameras, digital-audio players and recorders, and laptop computers support it. Peer closer, though, and you'll find the name Sony on almost all of these systems. Sony doesn't discuss the terms of its licensing proposals, so it's difficult to judge to what extent monetary considerations played a role in other vendors' lack of interest in Memory Stick.

Clearly, though, Sony is a formidable competitor in numerous application areas. (For example, Panasonic originated the SD Card in response to Memory Stick.) And Sony's competitors don't want to do anything that would add to Sony's coffers. Instead, other companies generally seem content to require that consumers purchase third-party adapters if, for example, they want to transfer pictures from a Sony digital camera to a non-Sony PC. Analogous to the evolutionary steps that MMC took, Memory Stick now comes in the Memory Stick Pro version, which has higher transfer rates, supports a larger directly accessible address range, and includes MagicGate encryption, and the Memory Stick Duo and Memory Stick Pro Duo versions, which have smaller form factors.

In the past, Fujifilm's and Olympus' cameras, along with early digital-audio players from companies such as Creative Labs and Rio (now Digital Networks North America), were the predominant users of SmartMedia cards. As that format faded from the limelight, Fujifilm and Olympus switched gears and migrated to a niche format of their own creation: the xD-Picture Card. Card vendors such as Lexar Media and Sandisk support this technology, which offers the predictable benefits: compactness and high-speed data transfers. Its primary advantage to Fujifilm and Olympus appears to be revenue-related; the companies' latest cameras work only with xD-Picture Cards, and proprietary formats as a general rule sell for two to three times more than their equivalent-density industry-standard alternatives.

The latest entry in the "I'll-go-it-alone; thank-you-very-much" sweepstakes is Motorola and Sandisk's TransFlash (formerly, T-Flash) module, which the companies intend for use in cellular phones. As with Memory Stick and the xD-Picture Card, it's unclear whether other system manufacturers will adopt TransFlash support. Unlike with the other formats, users will neither regularly remove TransFlash from the system nor interchange it with other systems, although, as with a cellular-phone-SIM (subscriber-identity-module) card, they can move it from one system to another. It provides the second of three possible scenarios for increasing the local-storage capacity in cellular phones, the first of which is the removable mini-SD and RS-MMC cards. The third option, which some GSM backers advocate, involves simply boosting the capacity of the SIM card currently in the phone.

In contrast to the proprietary and semiproprietary approaches of the other sections of this article, the USB-flash drive has been the beneficiary of built-in Windows operating-system support extending back to Windows ME. USB connectors first appeared in PCs in the late 1990s, and the more recent 480-Mbps USB High Speed variant keeps pace with the ever-increasing read and write performance of latest generation flash memories. As a result, the USB-flash drive, a product category that didn't even debut until November 2000, has rapidly grown to become a leading user of data-centric AND, NAND, and equivalent flash memory. People know it by various other marketing names, too, although the USB Flash Drive Alliance, with Sandisk noticeably absent from the membership list, is trying to solve that confusing problem.

With success comes inevitable product differentiation and fragmentation, although the commonality of the USB-hardware and -software interface has constrained any consequent incompatibility. USB-flash drives come in a plethora of shapes, sizes, and color schemes. Lexar Media's devices embed LEDs that indicate in-progress writes and help prevent users from blindly yanking the flash drive from its socket. Some USB-flash drives have rugged metal cases that withstand users' dropping, sitting on, stepping on, or driving over them. Others, according to their vendors, withstand extremes of heat and cold or immersion in fluids. USB-based devices that combine digital-audio-playback functions and flash memory for generic storage, not just music, are increasingly common, and a few USB-based still cameras and videocameras offering generic flash-memory storage also exist. Forget about a USB-flash drive on your key chain; wristwatches with flash memory inside and USB dongles for connectivity are here. USB memory is in pens, and Victorinox sells a Swiss Army Knife with a built-in USB-flash drive!

The software bundles that come with USB-flash drives are as diverse as the drives themselves. Some utilities will redirect a Web browser's cache, cookies, and other "fingerprints" away from the PC's hard drive, enabling you to anonymously surf with system diversity and impunity. Others perform similar flexible-system-usage functions with Outlook, Outlook Express, and other e-mail clients. Encryption schemes protect your precious data from prying eyes. And one-touch backup with on-the-fly compression enables valuable information to survive a hard-drive crash. Although the term "flash drive" describes this product category, at least one hard-drive manufacturer, Cornice, offers magnetic-media-based USB-storage devices with retail partner Digitalway. GS Magicstor, Hitachi, Seagate, Toshiba, and others that also offer and plan to offer 1-in. and smaller drives may provide the multisourcing muscle to broaden the hard drive's penetration in this market.

Microsoft, acutely aware of the ever-increasing popularity of USB-flash drives, demonstrated two intriguing scenarios at May's WinHEC (Windows Hardware Engineering Conference). In the first, a system booted a limited-function Windows XP variant from a USB-flash drive, a capability that's useful, for example, for IT administrators doing system debugging or upgrades. In the other, first plugging the USB-flash drive into a PC and then into other LAN devices, such as access points, routers, and printers, and then back into the PC enabled straightforward setup of a WiFi network. The demo employed encrypted storage and interchange of MAC (media-access-controller) addresses, SSIDs (Service Set Identifier), WEP (Wired Equivalent Privacy) or WPA (WiFi Protected Access) keys, DHCP (Dynamic Host Configuration Protocol) or static-IP details, and other relevant configuration data.

One USB spin-off that, at least at the moment, shows little likelihood of success is the oddly named Fish Memory that the UTMA (Universal Transportable Memory Association) promotes. Technical data on the association's Web site is sparse, and the association didn't respond to e-mails requesting details. The UTMA plans two sizes, Fish and BabyFish, but, without additional information, Fish Memory appears to be nothing more than a branded version of USB-flash drives that multiple suppliers already sell. After a brief flurry of promotional activity at February's PMA (Photo Marketing Association) show, the UTMA has gone quiet, and, at press time, no one had updated the Web site since Feb 10. No module or system vendors have announced backing for the format.

Many digital cameras and PDAs contain built-in USB and USB-derivative connectors, but their lack of USB Host controller capability precludes their direct attachment to USB-based mass-storage devices, such as flash drives, optical burners, and hard drives. (They instead embed reduced-function USB Client logic.) Delkin's dual-AA-battery-powered, $69.99 USB Bridge aims to surmount this limitation (Figure 2). Its support for 12-Mbps USB Full Speed mode minimizes cost but slows transfer rates to and from otherwise-speedier media.

Want to spend a bit more money to back up and transfer data from your diminutive memory cards using a more flexible platform but not interested in buying a full-featured and expensive laptop computer for this focused function? Kanguru and others will happily sell you devices that directly burn optical discs from memory-card contents, or transfer the data to a hard drive. You can also use the Kanguru Media X-change 2.0, which supports USB High Speed mode, as a portable USB-hard drive and multiformat-card reader; it comes with a built-in LCD and in 20- to 80-Gbyte capacities. The Kanguru FC-RW works with seven card formats, supports 36× multisession write speeds over High Speed USB, and doubles as a PC's CD-RW drive and card reader.

Multiformat card readers come in ATA, USB, and IEEE-1394 variants and either fit into a PC's floppy-drive slot or externally tether to the computer. Some adapters transform MMC, SD, Memory Stick, and xD-Picture Cards to the more prevalent CompactFlash and USB interfaces, whereas others translate CompactFlash and other card formats to a PC Card-compatible connector. Kanguru even includes supplemental flash-memory storage within its USB-based Micro Drive+ (for SD and MMC) and Kanguru Micro CF (for CompactFlash) card adapters.