The Milly Watt Project:

Preliminary Measurements of Power Consumption in PalmPilot

The following chart was developed to assess the cost of various power states in the Hiker's Buddy PalmPilot application. The measurements were done by developing an application that puts the Pilot into known power states and measuring current with a multimeter during each test period. We took the voltage values from the Pilot via sysBatteryInfo calls before and after the individual tests. A PalmPilot Professional running PalmOS version 2.0.4 was used.
Milly
Watt Proj
ect Icon

Bar chart of pilot power states

The results in the bar chart are averages of multiple experiments for each power state. We describe the details of the individual bars from bottom to top.

  1. The bottom bar shows the power consumed by a PalmPilot in sleep mode while sitting in its cradle connected to a PC. The .45mW for a sleeping device hardly shows at all to the left of the component representing the power consumed by being cabled to the PC.
  2. The next bar shows the power consumed by a PalmPilot in the idle state (e.g. waiting "forever" at the top of the event handling loop for the next user input event). In this power state, the display LCD is on, at medium contrast. The variation from maximum to minimum contrast spans a range from 39mW to 45mW in our experiments (the variation is not represented in the bar chart).
  3. The third bar from the bottom shows the power consumed by the PalmPilot in a busy state, looping through the event handling loop with a timeout of 1 clock tick, after which a nilEvent is generated. This represents a rather "tight" event handling loop with minimal processing involved. Typical event loops would fall between the idle value and this busy state.
  4. The next bar (4th from bottom) is for a computationally intensive busy loop with data in registers.
  5. The 5th bar is the power consumption of a loop involving memory reads and writes. These show the effect of doing significant amounts of processing.
  6. The 6th bar shows an idle CPU with the LCD's backlight on. The contribution of the backlight is shown as a separate component of the bar.
  7. The 7th bar shows the power consumption of doing continuous drawing with the stylus in the screen area.
  8. The next three bars show the effect of serial line operations such as those required for synchronizing data with the PC desktop application (hotsync in PalmPilot jargon). The first of these (the 8th bar) shows the power consumed by holding the serial line open to a PC. The individual contributions of CPU in idle state, simply being cabled to the PC, and having the serial line open are shown.
  9. The 9th bar shows the power cost of sending data over the serial line. These results were obtained by a controlled state of the test application in which data were sent to a PC running hyperterm. These results are based on sending a short record (7 bytes). Further tests with longer record lengths are in progress. The values were confirmed by measuring a hotsync operation dominated by sending.
  10. The 10th bar shows the effect of receiving over the serial line. A hotsync operation, this time dominated by data being received, was measured. The transfer consisted of installing a 200K doc database.
  11. The next two bars show power states that are relevant for the Hiker's Buddy application in which the serial port is connected to a GPS receiver. Since our GPS unit is turned on and off by opening and closing the serial line, a state of idle CPU and serial line open to GPS is important as the cost of maintaining GPS state between requests for a location fix by the user.
  12. The topmost bar shows the state of actively receiving NMEA protocal sentences from the GPS unit in response to a user's location request.

Other measurements of Pilot power consumption have been done. In particular, see http://www.massena.com/darrin/pilot/luiz/item7.htm .


Carla Ellis's HomePage | Mobile Computing Research