在E-Prime程序中如何提高时间精度——操作部分张豹 2006-2-22提高和检验时间精度的六个步骤：步骤一、测试和调整实验所用的计算机的时间如果计算机的硬件不支持毫秒的精度,软件在运行实验程序时就会产生误差.所以在运行程序之前必须测试计算机是否支持毫秒级的精度.E-Prime提供了测试程序来确定计算机是否支持毫秒级的精度.[refer to Appendix A: Timing TestResultspage 217]当计算机安装新的硬件或是软件时有必要运行时间测试程序,时间测试能保证桌面PC硬件为Pentium, 120MHz 或更高时E-Prime会达到毫秒级的精度. 当实验用键盘或PST反应盒外部设备进行反应时能达到毫秒级,但是一有般不推荐作鼠标作为E-Prime实验程序的反应装置来保证毫秒级的精度.[Using a quality name brand PCI audio card, E-Prime can achieve consistent low latency playback of digital audio . We do not recommend using a legacy ISA audio card for experiments that require optimal and consistent playback latency. ]一个好的实验时计算机必须有最小数量的程序同时运行,在数据收集时没有其他的应用程序程序运行.理想状态是在收集数据的时候除了E-Run之外在窗口工具栏中没有其他任务程序运行,而是运行背景程序〔其他操作系统的程序〕也应最少.〔按Ctrl+Alt+Del可以显示当前运行的程序〕运行刷新时钟测试〔RefreshClockTest〕PST提供有测试软件来收集和分析在WINDOWS95/98/ME的计算机上的时间数据,目的是评估研究所用的计算机的时间精度,可以从上下载.测试实验的步骤：1. Download and unzip the test .下载软件2. Close all other open applications .关闭其他的应用程序3. Launch E-Studiofrom the E-Prime menu via the Start menu.打开E-Studio程序4. Open RefeshClockTest.ES in E-Studio.在E-Studio中打开的RefeshClockTest.ES.5. Generate and run the experiment by pressing the {F7} key. When E-Prime promptsfor Subject, Session, and Group values, it is suggested that a machine ID be enteredas the Subject number, the sequential run number be entered as the Session number,and the test ID be entered as the Group number. It is also suggested thatthe user keep a written log of the results.运行程序6. Read through the prompts and then select the duration of the test: Enter "1" to run thetest for 1 minute, "2" to run for 25 minutes, or "3" to run the test for 6 hours.输入运行时间,1－1分钟,2－25分钟,3－6小时.E-Prime提供了很多时间精度的测试,但最重要的是CLOCK TEST,时钟测试连续监控计算机时钟10000ms.E-Prime能够检测到读取时钟时是否有中止以跳过clock tick〔例如值是2001,2002,2005,2006,这表明在2002,2005之间有3ms的tick〕,这表明在实验过程中操作系统有时间误差.另一个测试是Refresh Test,这个测试有三个功能：1,检测显卡硬件是否提供垂直空格信号〔vertical blanking signal〕〔例如在刷新时光栅从屏幕底部到顶部时的时间〕.2,检测硬件的missed refreshes的速度,有的显卡的垂直空白之间呈现的信号很短〔10ms〕,这样短的信号不能被E-Prime稳定地检测到,而要维持刺激与显卡的同步必须要检测到垂直空格事件.因此,为了保持同步,在不用垂直空格模拟编码时missed refreshes的速度的检测必须小于0.10％,在用垂直空格模拟编码时要小于50％.影响时间精度测试不合格的因素：1. Failure to exit all programs before the test.其他应用程序2. Virus-scan programs and background programs 病毒软件以与背景程序3. Network traffic or shared disk activity interrupting the computer.网络或共享4. A slow machine that is unable to provide good timing.机器太慢不能提供好的时间为了达到稳定的结果,要关闭所有的其他应用程序〔亦可包括杀毒软件和网络,共享等〕,重新运行时间测试,运行一分钟的直到稳定,然后再是二十五分钟的,再是六个小时的.绿＋代表达到毫秒级；红×代表没有达到；黄色？代表边缘达到.通常第一次测试结果较差这是由于启动测试程序操作系统不稳定测试结果解释：缺失的刷新周期,比如没有检测到的垂直空格信号的启动的次数.如果太太,则重新设置屏幕参数〔分辨率,颜色深度,刷新周期〕,或更换显卡.需要分析的时间数据：TicksMissedPercent.ANLShows the missed tick percentage by test run. This value should be below 0.1%. You may seea larger number of misses on the initial test runs as the operating system finishes loading andinitializing the test software. If this is the only problem, it can usually be overcome by running ashort block of practice at the beginning of the experiment. If you have large spikes duringsubsequent runs, it indicates that the operating system, other applications, drivers, or hardwareare taking execution cycles from the experiment. You should attempt to review the state of yoursystem to identify other programs or hardware that may be causing delays in the experiment.用来显示missed tick的百分比.如果在几次测试之后还是有比较大的峰,这表明有其他的应用程序,或其他原因造成在实验时不能达到毫秒级精度.TicksMaxMissed.ANLShows the maximum missed tick duration by test run. This provides a good measure of theworst-case time delays caused by the operating system, other applications, drivers, orhardware. It should be < 10ms.BinCount.ANLShows the distribution of missed ticks across all test runs. Each bin holds the count of observeddelays of a specified duration. Bin #1 is the count of the detected ticks <1ms>. Bins 2-20 are thecounts of greater delays . Bin 21 is the count of all delays greater than 20ms. Bin22 is the count of any delay that was negative in duration . You can use thebin profile to determine if other programs or hardware are consuming time.RefreshMissedPercent.ANLShows the percentage of missed refresh cycles . This valueshould be less than 0.10%. If it is higher, the vertical blank simulation feature of E-Prime shouldbe used . If the rate is > 50%, you should re-run the testusing different display parameters to see if the resultscan be improved. If the rate cannot be reduced below 50%, you should consider changing thevideo display adapter or possibly not using the computer for experiments requiring shortduration, precise visual stimulus presentations.E-Prime时间测试〔E-Prime Timing Tests〕〔page 227〕对于一个研究实验,时间必须进行严格地测试.PST已经提供了一种自动地时间测试。
