Miniature disk drives in handhelds demand creative power management

But such capacity is not without its engineering penalties — especially for the power-delivery system. The next-generation handheld device — typically a cell phone, doubling as pocket media center — must provide surround sound audio (with at least two miniature speakers mounted inside the phone), and high-resolution, 30-frames-per-second video. It must capture 3-Mpixel photographic images, in the same way as a digital still camera; that is, with an autofocus mechanism and high-intensity flash. The portable must all the while maintain contact with a radio tower (a cellular basestation), for the dual purpose of downloading Internet pages and making voice calls. How anyone gets more than 15 minutes of battery life out such a device is either a small mystery or a major tribute to technology. But to the already-complicated design issue of squeezing battery life out of a feature-laden handheld, we add the problem of driving miniature motors.

The power management and delivery chain for handheld cameraphones is fairly sophisticated and complex, involving as many as 17 separate voltage regulators. The use of micro-miniature HDDs requires additional voltage regulation. The parts must be capable of delivering a relatively large amount of current on short notice (in burst mode), a smaller amount of current as the disk spins, and to come offline and go to sleep (certainly, consume no current of their own) when the drive needs to stop.

Click to Enlarge

The key to extended battery life with feature-laden handhelds, is in the voltage regulator topologies, and the efficiencies associated with each topology. Switch-mode regulators have come to replace linear low-dropout regulators (LDOs) in a wide variety low-voltage circuits. Their higher efficiency contributes not only to lower heat dissipation in confined spaces, but also smaller-sized filter components (inductors and capacitors). But switchers have certain drawbacks when used in battery-powered handhelds like cell phones and MP3 music players: Because switchers will use a pulse-width modulator (PWM) to rapidly toggle FET switches on-and-off, they can push switching noise into sensitive circuits. Additionally, switchers have had a tendency to lose efficiency under light current loads.

Thus, it's a challenge to the engineering talent of voltage regulator manufacturers to come up with devices which satisfy the particular requirements of the HDD-based cell phone or music player. There's a kind of genius at work here, in the power management device's ability to deliver just the voltage-and-current you need, but otherwise get out of your way.

Examination of the micro-miniature drives available for handhelds offer a relatively small number of choices. One is manufactured by Toshiba; another is made by Hitachi (with technology inherited from IBM). A startup called Cornice (Longmont, CO) claims to have shipped one-million a 1-inch MP3 "storage elements" to companies like Philips and Samsung. Like other HDDs, this one has an IDE interface and stores up to 3 Gbytes.

The Toshiba hard drive used in the original Apple iPod was a 30-GByte drive, with a 1.8-inch platter and 15-millisecond seek time. It demands a 3.3-volt power rail and draws up to 480 mA on startup. (Toshiba specs say "1.3 watts" —400 mA — during "read-write" cycles, 0.23 watts during standby and 0.05 watts in sleep mode.) A teardown report by David Carey of Portelligent Inc. (Austin, Tex.) suggests that the drive is buffered by 32 Mbytes of SDRAM. That is about 20 minutes-worth of skip-free music, he estimates.

The smaller Apple iPod "Mini" can store approximately 1,000 songs, using a 1-inch diameter 4-GByte Microdrive manufactured by Hitachi. (Some of the newer music players are making use of the Toshiba 0.85-inch diameter drive, as well.)

System memory in iPod Mini, Portelligent reports, includes 32 MBytes of SDRAM made by Hynix and 1 Mbytes of NOR Flash from Silicon Storage Technology. The audio codec made by Wolfson Microelectronics (Edinburgh, Scotland) integrates a headphone amplifier (the WM8731). A FireWire transceiver (the TSB41AB1) comes from Texas Instruments (Dallas).

For power management, the system has a USB Power Manager & Li-Ion Battery Charger manufactured by Linear Technology Corp. (Milpitas, Calif.) (the LTC4055), a hysteretic buck controller manufactured by National Semiconductor (Santa Clara, Calif.) (the LM3485), a step-down converter (the TPS62046) manufactured by TI, and a Power & Battery Management Controller (the PCF50605HN) manufactured by Philips Semiconductor (Sunnyvale, Calif.), Portelligent says.

The Li-Ion battery (3.8 V, 630mAh) is claimed to deliver up to 8 hours of continuous playback, with an fast-charge time (to 80 percent capacity) of one hour and a full charge time of 3 hours.

The biggest power management issue for a disk-based music player like the iPod Mini is in responding efficiently to the current spikes demanded by the disk drive as it starts to spin. In the power consumption profile assembled by Portelligent, there is roughly a 5-second period in which power consumption will rise to 1.5 watts as a song is being accessed for play from the disk. During the actual playback, the power consumption drops to 0.4 or 0.2 watts, depending on whether the user keeps the backlight while the song is playing. Power will go up and down within a 0.5-watt range, depending on the user's application of the iPod's click wheel and display — but the drive accesses and seeks clearly generate the largest power consumption spikes, according to Portelligent's report.

Figure 2. Power consumption curves for the Apple iPod Mini as a function of disk drive activity.

Click to Enlarge

The buck-boost is back

Some engineers may remember the debate from cell phones: The Li-Ion battery used by a cell phone maker like Nokia has a nominal 3.6V level. Much of the internal logic of the handheld device runs off of a 3.3- or 3.0-V rail. Even with the move to lower processor voltages (2.8, 2.5, or lower), much of the cell phone peripherals have required 3.3V. Thus, a series of step-down regulators are required to provide a controlled 3.0-V supply rail from a 3.6-V battery source. Some manufacturers recommended LDOs to ensure low noise; others perfected step-down switching regulators using a "buck" topology.

It remained uncertain, with early generation cell phones, how much battery capacity was left after the battery voltage drooped below 3.0 or 3.3 V in sustained use. Continued use would cause the measured battery voltage to droop below 3V, and some estimates suggested there could still be as much as 10 percent of the battery's capacity still left at 3.0V. Some manufacturers argued that a step-up topology — a "boost" regulator — might be appropriate to lift the supply rail voltage back up to 3.3V once the battery voltage itself drops to 3.0 or 2.8.

While buck regulators and boost regulators can offer high conversion efficiency, the type of regulators that start out as a buck, and then convert themselves to a boost — a "buck-boost" — might be too inefficient to sensibility extract a little extra current from a Li-Ion battery that's nearly depleted. A Nokia might just sacrifice a small amount of talk time and turn the cell phone off, opined Dave Heacock, general manager of Texas Instruments' portable power management business unit in Dallas, some five years ago.

The European cell phone makers still have a tendency to turn off the phone when the battery voltage drops below 3.0 or 3.1 V, says Tony Armstrong, portable product marketing manager with Linear Technology Corp. However, Korean cell phone makers will let the voltage go down as far a 2.8 or 2.7 volts, he believes. "You don't want to pull so much current at lower voltages that you damage the battery," Armstrong reminds.

But the choice to shut off the handheld at 3.0 is no longer an option for designers of disk-based handheld music players. While the internal logic of the player, even the peripherals and I/O circuits, now run on significantly lower voltages, the HDD still needs 3.3. Thus, voltage regulator manufacturers like Linear Technology have come up with new buck-boost topologies — even devices that will supply the burst-mode currents typically demanded by a disk drive seek operation.

One example is LTC's own LTC3442, which delivers up to 1.2 A at 3.3V, and offers up to 95 percent efficiency. The device takes its input voltage from a Li-Ion battery, and maintains a 3.3V output, even as the input degrades from 4.2 to 2.5 V. The switching frequency of the LTC3442 is adjustable from 300kHz to 2MHz, it has the necessary MOSFET drivers integrated on chip.

Having acquired additional sophistication with battery fuel-gauging, Texas Instruments is now re-considering the actual power available in a nearly-depleted Li-Ion battery. "Portable power management is no more a function of transient response, than auto fuel economy is strictly a function spark plug ignition," Dave Heacock now comments. "You have to consider the entire power delivery chain." New battery chemistries are allowing enable more energies to be extracted at low voltages, he says. (See, for example, "Battery topology holds key to efficiency.")

On the horizon, though, are 4.2 and 4.4V Li-Ion batteries that offer as much as 30 percent more capacity. "Sanyo is off to the races," says Tony Armstrong.

Such higher-voltage sources might eliminate the need for boost conversion, allowing an HDD to be powered entirely by a step-down regulator. Micrel Inc. (San Jose) has given considerable thought to the issue of driving micro-miniature HDDs, and is proposing two solutions. One is a family of synchronous buck converters (the MIC2202 and MIC2205) that transform themselves to from switchers to LDOs under light current loads.

The other is a device, under development, powers HDD-based through their USB interface. Though USB limits current to 500 mA or below, explains Thurston Awalt, product marketing manager, "a 'kick-start' feature on USB controllers offers a 125-ms window to start the drive in motion [before current-limiting sets in]." The 40-80 mA spin cycle could then be accommodated through the USB interface, he says.

Figure 3 A scheme for powering HDDs current spikes from USB.

Click to Enlarge

"It takes 3-4 minutes read a song out to the Flash," confirms iPod fan Bernie Weir, engineering manager for On Semiconductor (Phoenix). The bursty nature of the drive activity, combined with the 3.3V requirement, may require a buck-boost, he agrees. The closest that On Semi comes to this are the NCP1510A and NCP1511 buck converters, designed to deliver up to 500 mA at 3.3V from a Li-Ion battery.

Getting inside the drive

Much of what we know about driving disk drive motors might be extrapolated from our experience in driving shutters and auto-focus mechanisms in digital still camera, suggests Nazzareno "Reno" Rossetti, director of IC strategy at Fairchild Semiconductor Corp. in San Jose. Fairchild, for one, offers a series of miniature H-Bridge drivers with bipolar transistors (see, for example, "Digital cameras make use of integrated motor drivers"). Designers of music players can similarly come up with efficient schemes for driving motors inside disk drives — if indeed the drive mechanisms were accessible.

How efficiently we can power a miniature hard disk drive (HDD) is somewhat contingent on how much access we have to the innards of the drive itself. There are two motors here: one to spin the magnetic disk; the other (more of a voice-coil actuator) to position the read-write over the appropriate track on the disk. There is also a hard drive controller (HDC) which determines where everything is parked on the surface of the disk and the sequence in which it is retrieved. Parking and retrieving digitized music and video on a disk may be a somewhat different than (say) a Microsoft Word file.

To store a Word file, the HDC could scatter its pieces all over the surface of a disk. The user probably doesn't mind staring at an hourglass while the HDC retrieves the various fragments and reassembles the file. The iPod user would rather have his music interrupted by an hourglass. In operation, the iPod music player will gather all fragments of a song, and reassemble it in flash memory before beginning the playback sequence. But the process of finding and assembling digital music fragments could be streamlined — and use less power — if the music were parked in contiguous tracks and sectors on the drive.

"How you manage data in these applications is key," says Dan Fisher, vice president for system-on-chip R&D at STMicroelectronics' data storage group, whose company manufacturers read channels, HDC systems-on-chip and motor drives for hard disk drives. "You can save energy by avoiding track changes," Fisher says.

Fisher has looked for object-oriented software to help control the head-cylinder interaction in the drive, particularly the head boundary, and the track boundary. For MP3 music players, there is a tradeoff between "dumb storage" and "intelligent storage." A true consumer operating system would lobby for "dumb storage," Fisher claims, in order to get greater control over the head-cylinder interface.

It turns out, the same device that is used as an "applications processor" in the Apple iPod, as well as in SigmaTel (Austin, Tex.) nearly-ubiquitous MP3 SoC-codecs — the ARM9 — is also the HDC processor which controls the seek-and-play functions on a wide variety of HDDs.

"Managing power — what's turned on and what isn't — would be much easier if the applications processor and the HDC were in direct communication with each other," says Neil Robinson, stateside marketing director for the ARM Consortium (Los Gatos, Calif.) "The ATA interface requires a lot pins and sucks a lot of power," he agrees. In a different partitioning scheme, the HDD absorbs the applications processor — perhaps with a dual core ARM devices — and the ATA interface goes away.

The trend is toward moving HDC control functions toward the applications processor. That gives the music player people more control, but it is not a power saver, Robinson believes. It simplifies the drive, but complicates the applications processing tasks. "It should be the other way around," he says.

The drive makers are not likely to allow it, Robinson concludes. A dumb drive offers more versatility. "The Apples of this world will still need to pick-and-choose."