Power pinching. (reducing power requirements for laptop computers) (Column)
by Mark Minasi

In the last two columns, I've looked at laptop battery technologies and suggested some ways to make batteries last longer. But there's a limit to the amount of power that engineers can pack into a battery. Extending laptop life means using less power somehow.

One question I get when I talk about this subject concerns solar power. "Isn't it possible to power a laptop with solar power?" people sometimes ask. The answer is, "Not yet, but eventually." There are solar panels for laptops that can provide about 500 milliamps of power; notebooks need around 3000 milliamps. You'd need a pretty big set of panels to power a laptop.

There's not much we can do about solar-power research, so let's consider a second approach--lowering laptop power requirements. In descending order, the biggest users of juice in your laptop are the display, the hard disk, the RAM, the floppy drive, the CPU and support chips, the keyboard, the system clock, the I/O ports, and the math coprocessor (if present). Let's examine these power porkers in order.

Displays draw the most power. I see that every time I connect my laptop to an external display--the laptop runs for hours and hours off a single charge. Displays would be a lot easier on the batteries if they were (1) slower, (2) lower resolution, and (3) not backlit. Displays must be refreshed many times per second, meaning that an electrical circuit must repaint the image on the LCD screen usually about 50 times per second.

Here's a side note that will be useful for the rest of this article. Any circuit that does things on a regular basis, like a clock ticking or a video circuit refreshing a screen, is an alternating current (AC) signal. The frequency of that signal affects the circuit's power-consumption rate like so: P=CV[sup.2].

In that formula, C refers to the Capacitance of the circuit, V refers to the Voltage of the circuit, and f refers to the frequency of the circuit. In terms relevant to our current problem--displaying data on a computer screen--the term frequency relates to the video refresh rate. Typically, a screen should refresh at about 60 times per second, but 50 is acceptable, and look what it does for the power consumption. Changing the frequency from 60 to 50 reduces display power consumption by 17 percent.

Unfortunately, that refresh clock isn't driven by the number of screens painted per second; it's driven by the number of lines painted on the screen per second. Suppose we're building a CGA-compatible display; CGA screens have 200 horizontal lines on them (as you may know, CGA resolution is 640 x 200). This means that the line clock must tick 200 x 50 times per second. But today's notebooks have at least a VGA resolution, and VGA has 480 lines of horizontal resolution. This means that going from CGA to VGA resolution increases laptop screen power requirements by 240 percent!

The next laptop display problem comes from backlighting. Supertwist LCDs show high resolution, but they really need to be backlit to be visible (in my opinion, that is--there are certainly supertwist LCDs that aren't backlit, but I find them unreadable). The fluorescent light behind a backlit LCD draws substantial juice.

What can be done to keep laptop power consumption down? You can shut down the laptop display altogether during inactivity. It's annoying, but if you're like me, you probably get distracted in the middle of battery-powered work by airline seatmates, flight attendants, or the like. Those extra minutes of display down-time can significantly extend battery life. And whenever possible, turn the display brightness down, reducing the amount of power that the fluorescent tube needs.

Today's laptops can't live without hard disks, as evidenced by the fact that you just can't run Windows from floppies. Hard disks have a motor that keeps the disk platter spinning (as well as a voice-coil circuit that moves the read/write head) in addition to having to power the electronics on the hard disk itself.

The really big disk amp-sucker, however, is the action of powering up the hard disk. It takes a lot less power to keep a disk spinning than it does to get it spinning in the first place, which makes me leery about the common laptop practice of shutting down the hard disk when inactive. My laptop came out of the box set to shut down the hard disk after one minute of inactivity, a setting that led to near-constant power ups and downs for the hard disk. Not only does that draw power, but it's just plain no good for the hard disk motor, and surely shortens its life. Consequently, I compromised and set my disk timeout to the maximum allowed by my setup program, 15 minutes.

An interesting bit of good news for laptops comes from an examination of what determines how much power is required to get a disk spinning in the first place. The amount of power needed to get a disk spinning is proportional to the cube of the radius of the disk. If that doesn't seem interesting, consider this--notebook hard disks used to have platter diameters of 3 1/2 inches, but most now use a 2-inch-diameter platter. This means that the startup power required for a 2-inch disk is smaller by a factor of (3.5/2) cubed, or over 530 percent! 1 7/8-inch platters are starting to appear, which will mean even lower-power-consumption rates.

Buy laptops with small platters where possible. Don't worry so much about capacity, as that's not as important in power consumption, and from a practical point of view, you should have as much disk space as possible in order to run today's software.

Experiment a bit to find out the best timeout value for your laptop's hard disk. You don't ever want the hard disk to power down while you're scratching your head looking for the right word. Instead, you want the laptop to power down when you've turned away to answer the phone or chat with the person sitting next to you on the airplane.

The next power purloiner is the system's main memory. My notebook must have 16MB of RAM, as I run Windows NT on it. But more memory needs more power . . . usually. While it depends on how the memory is laid out on your laptop, you'll typically find that the fewer the number of memory chips in your system, the lower the power consumption. For example, suppose you have a laptop that can accommodate eight SIMMs (Single In-line Memory Modules). (These are small circuit boards about the size of a stick of gum that are the typical packaging for memory these days). You need 8MB of RAM, and you can either get those 8MB with eight 1MB SIMMs or two 4MB SIMMs. The two 4MB SIMMs will use significantly less power.

Be careful when you're buying those SIMMs. Use the SIMMs that the manufacturer recommends, or you'll throw away power like crazy. A lot of what makes memory draw power is that it must be refreshed on a regular basis. Think of each memory location as being like a small storage container for colored water. Red colored water represents a 0, and blue colored water represents a 1. A program stores data by putting water of the appropriate color into a container. So far, so good.

Unfortunately, the storage containers are leaky. Many times per second, your memory chips must be refreshed. What I've just described here is a dynamic RAM. The alternative is a static RAM, which is essentially composed of containers that don't leak. You put data in the containers, keep a steady supply of power to those containers, and the data stays there. (Remove the power, and even a static RAM loses its contents.)

The refreshing that the dynamic RAM requires is accomplished by a circuit that constantly rereads the memory, with the resultant side effect that the storage containers get refilled. That process requires a significant amount of power, so many notebooks are built around low-refresh SIMMs, which are memory modules that don't need to be refreshed as often as normal SIMMs. Buy normal SIMMs, and your notebook will run fine, but you'll seriously degrade its battery life. So check with your manufacturer before you buy upgrade memory for your system.

There's not much that can be done about the floppy, as it doesn't require power unless you're accessing it. But what about the CPU and support circuitry? Look once again at the formula relating AC circuits to power consumption. The entire motherboard of your laptop moves to the beat of the central system clock. A clock that runs at 20 MHz has a frequency of 20 million cycles per second, and you recall that the higher the frequency, the higher the power drain. Smart laptops detect idle time and drop the clock or, even better, stop the clock altogether. It's not quite as easy as that, however, as many of the memory components inside the CPU itself are dynamic and require refreshing. You can't just stop the clock on many motherboards and expect the data to remain intact. That's why Intel developed the 386SL, a processor basically intended for laptops. You can stop its clock without any trouble.

To see another way to save power, consider this power-consumption formula for direct current (DC): P=[V.sup.2]/R

Here, Power consumption equals Voltage squared divided by Resistance. Traditionally, chips have run at 5 volts. Some manufacturers reduce voltage by just running the chips at a voltage level a bit below their rated values. For example, a circuit that uses three AA batteries would only have 4.5 volts, leading to a nearly 20-percent savings in power in the circuit. Another approach is to use a chip that draws less power. Intel has a 3.3-volt version of the 386SL available, but it hasn't really caught on because vendors seem to be waiting for the 3.3-volt version of the 486.

I'm running out of space, so here are a few more ideas for the lesser power drains. Many notebooks let you disable the serial ports with the setup program that comes with the notebook. As I only need one serial port for my mouse, I disable the other serial port and the parallel port. If you can live without a math coprocessor, do so, as it runs a bit warm, and you know where the power for that heat comes from--your battery.