功率消耗量對於嵌入式系統平台是有用以及重要的資訊，我們可以使用功率消耗資訊來省電以及延長系統平台的使用時間。完整的系統耗電量描述包括硬體元件的耗電以及應用程式執行行為所造成的耗電。以往的系統耗電工具多為模擬，即以處理器的執行計數器搭配功率模型評估系統。然而，使用模擬方式並不能滿足所有的實際狀態。因此，我們實作了兩個可以量測不同硬體規格的印刷電路板，第一版的功率量測硬體是由乙組數位轉類比積體電路，八組電流旁路監視積體電路以及八組跨壓電組所組成，第一版可以量測Android 待機以及開機的功率消耗。第二版的功率量測硬體是由八組數位轉類比積體電路，八組電流旁路監視積體電路以及八組跨壓電組所組成。第二版量測的解析度依據不同的裝置分別為277.433uA 以及55.4865uA。其中第二版可以透過FPGA 來調整數位轉類比的取樣頻率，可以避免取樣頻率的問題，非常的有彈性。在實驗結果部分，我們用功率量測模組跟安捷倫示波器 (Agilent DSO6052A)進行比對驗證，了解功率量測模組的精準度。ARM core 執行立方方程式最大誤差為3.49%，執行角度換算程式3.73%，執行傅立葉轉換程式3.44%，LCD 最大誤差為3.98%，GSM 最大誤差為3.25%，記憶體模組最大誤差為5.01%，無線網路網卡最大誤差為8.53%。ARM core 執行立方方程式平均誤差為1.89%，執行角度換算程式1.85%，執行傅立葉轉換程式2.06%，LCD 平均誤差為1.11%，GSM平均誤差為0.88%，記憶體模組平均誤差為1.58%，無線網路網卡平均誤差為1.4%。除了ARM core 以外大部分的誤差都分佈在小於1%，ARM core 分佈在1%~2%以及2%~3%之間。

Power consumption of embedded system platforms is an important and useful information, we
can use power consumption information to save power and extend using time of system platform.
The complete system’s power consumption is described by the hardware component’s power consumption
and the application program execution behavior. In the past, the tool of system power
consumption was simulation-based. It was using the performance counter of processor and the
power estimation model. However, it can not satisfy all real status. Hence, we have made two
hardware measurement modules, it named power measurement board version 1 (PMBV1) and version
2 (PMBV2); it can measure different hardware of specifications. The PMBV1 consists of
1 ADC IC (ADS7844), 8 current shunt monitor ICs (INA195), and 8 resistances, it can measure
Android idle and booting. The PMBV2 consists of 8 ADC ICs (ADS805), 8 current shunt monitor
ICs (INA19x), and 8 resistances. The PMBV2’s measurement resolution is 277.433uA and
55.4865uA depend on different device. The PMBV2 can be adjusted ADC’s sample rate through
FPGA, it can avoid the sample rate issue and it is very flexible.
In experimental results, we have verified and compared measurement result with oscilloscope
(Agilent DSO6052A) and power measurement board to realize the accuracy of power measurement
board. The maximum error of ARM core of benchmark cubic function is 3.49%, ARM core of
benchmark angle convert is 3.73%, ARM core of benchmark FFT is 3.44%, LCD is 3.98%, GSM
module is 3.25%, memory module is 5.01%, and USB WiFi is 8.53%. The average error of ARM
core of benchmark cubic function is 1.89%,ARM core of benchmark angle convert is 1.85%,ARM
core of benchmark FFT is 2.06%, LCD is 1.11%, GSM module is 0.88%, memory module is
1.58%, and USB WiFi is 1.4%. The most errors of devices are distribution in 1%, besides ARM
core is distribution in 1%2% and 2%3%

[1] Y. Cho, Y. Kim, S. Y. Park, and N. Chang, “System-level power estimation using an on-chip
bus performance monitoring unit,” in Proceedings of IEEE/ACM International Conference
on Computer Aided Design (ICCAD), November 2008, pp. 149–154.
[2] Y.-L. C. LIANG-BI CHEN and I.-J. HUANG, “A real-time power analysis platform for
power-aware embedded system developments,” 2009.
[3] Atmel, AT91CAP9A-DK Development Kit User Guide, 2007. [Online]. Available:
http://www.atmel.com/dyn/resources/prod documents/doc8506.pdf
[4] ——, AT91CAP9 MPBlock User Guide (Summary), 2007. [Online]. Available: http:
//www.atmel.com/dyn/resources/prod documents/6324s.pdf
[5] T. I. (T.I.), ADS7844, 2003. [Online]. Available: http://www.ti.com/lit/gpn/ADS7844
[6] ——, ADS805, 2002. [Online]. Available: http://www.ti.com/lit/gpn/ADS805
[7] Atmel, AT91CAP9S500A/AT91CAP9S250A, 2009. [Online]. Available: http://www.atmel.
com/dyn/resources/prod documents/doc6264.pdf
[8] T. I. (T.I.), INA195, 2006. [Online]. Available: http://www.ti.com/lit/gpn/ina195
[9] Hitachi, TX09D70VM1CCA Datasheet, 2008. [Online]. Available: www.hitachi-displays-eu.
com/doc/TX09D70VM1CCA.pdf
[10] T. I. (T.I.), INA193, 2006. [Online]. Available: http://www.ti.com/lit/gpn/ina193
60
[11] Siemens, MC39i Datasheet, 2005. [Online]. Available: https://www.swe.siemens.com/italy/
web/AD/ProdottieSoluzioni/HomeSC/ModuliWirelesseSoluzionim2m/PortafolioProdotti/
Documents/CP-AD/Sito/SC/WM/PortafolioProdotti/Datenblatt MC39i20GB 1189266.pdf
[12] KEITHLEY, KEITHLEY 2400, 2005. [Online]. Available: http://www.keithley.com.tw/
products/localizedproducts/sourcemeter/2400
[13] M. Guthaus, J. Ringenberg, D. Ernst, T. Austin, T. Mudge, and R. Brown, “Mibench: A
free, commercially representative embedded benchmark suite,” Workload Characterization,
Annual IEEE International Workshop, vol. 0, pp. 3–14, 2001.
[14] Atmel, AT91CAP9 SAM-BA User Guide, 2010. [Online]. Available: http://www.atmel.com/
dyn/resources/prod documents/6421B.pdf
61