Stimulus Presentation Challenges
|The hardware that presents stimuli, whether visual (video card, monitor) or auditory (sound card and its related drivers), restricts the rate at which stimuli can be presented.|
|Visual Stimulus Limitations:|
• Displays are accomplished via refresh cycles.
• Refresh cycles are rarely exact multiples of the intended display duration.
Auditory Stimulus Limitations
NOTE:The hardware that comprises the computer display, along with the process by which monitors display information, determines the display durations that are possible.
Therefore, in order to understand the limits on display durations, we need to examine some of the details of the hardware and how it works. Computer display devices, both CRT and LCD displays, paint every dot or “pixel” of a display sequentially. The display begins to be drawn at the top left, and pixels are then painted horizontally across the row, move down to the second line and then continue horizontally. The process is repeated for each row of pixels until the bottom of the screen is reached. This process of painting all the pixels sequentially for a full screen is called the refresh cycle, and typically takes 10-18 ms (the “refresh duration”).
The start of a refresh cycle is referred to as the vertical blank event. On a CRT monitor, this is when the electronic gun on the monitor moves from the bottom right of the display to the top left of the screen to restart the refresh. The computer can sense the vertical blank event in order to synchronize the computer events to what is occurring on the display. The display card, the monitor, and the screen settings (e.g., resolution and color depth) all determine the refresh cycle or refresh rate.
LCD monitors also have a vertical blank event. However, in contrast to the CRT monitor, the vertical blank event does not refer to a physical event, because an LCD monitor does not have an electron gun. LCD monitors do, however, have a delay after the full screen has been “painted”, during which information about the next screen to be displayed is transferred from the video buffers of the hardware card itself and then through the analog or digital connection to the LCD. This data transfer time imposes a limit on how quickly the display can be updated, similar to the limitation imposed by the vertical blank event on a CRT. Largely for purposes of backwards compatibility with CRT based systems, this period of data transfer from memory buffers to the video card is still referred to as the vertical blank event.
A typical display resolution includes 1024 pixels horizontally and 768 pixels vertically. Figure 1 shows time progression of the display of each pixel row in a 1024x768 display when a 60 Hz refresh rate is in effect.
Figure 1. Video display refresh cycle for a 1024x768 display with a refresh rate of 60Hz (16.666 ms refresh duration). The top line of the display is line 0, and the bottom line is 767. Each pixel is displayed once every 16 ms.
To calculate the refresh duration in milliseconds given the refresh rate specified in Hz, use the following equation:
Refresh Duration (ms) = 1000 / Refresh Rate (Hz)
Example: Refresh Duration = 1000 / 60
Refresh Duration = 16.666 ms
To calculate the refresh rate in Hz given the refresh duration specified in milliseconds, use the following equation:
Refresh Rate (Hz) = 1000 / Refresh Duration (ms)
Example: Refresh Duration = 1000 / 16.666
Refresh Duration = 60Hz
On a CRT monitor, during the refresh cycle the video display hardware activates each pixel for a short period of time (about 3 ms per refresh cycle). Figure 2 shows the typical activation the eye sees for a pixel in the middle of the display. The example presents a fixation, probe, mask sequence of “+”, “CAT”, and “***” respectively. Each write of the data to the video hardware is assumed to occur during a vertical blanking interval (e.g., screen writes performed at offsets of 0, 16 and 32 ms). The pixels for the middle of the display (line 384) would be refreshed and displayed at the approximate offsets of 8, 24, and 40 ms.
Figure 2. Pixel activation of center of display for “+” presented at 0 ms, “CAT” presented at 16 ms, and “***” presented at 32 ms. The participant actually first sees “+” at 8 ms, “CAT” at 24 ms, and “***” at 40 ms
The arrows at the bottom of the figure illustrate when the text is written to the video display card. The exponential decay functions show what is actually displayed to the participant’s eyes when the experiment is presented using a CRT monitor. The delay in the activation is due to the vertical refresh cycle. The intensity change is due to the phosphor decay of the CRT monitor after the pixel was activated once per refresh cycle. Our testing shows that the phosphor decay is approximately 1.2 ms, and further that it is fairly consistent across different monitors.
LCD monitors do not experience the phosphor decay effect because of the technology they use to maintain the image. As noted above, LCD monitors are still participant to the concept of a refresh cycle to update image data on screen, but once updated the image is does not decay. Instead, it is rather continuously illuminated via florescent light or similar technology.
However, LCD monitors introduce a new timing issue, display latency. Display latency is the delay between when the monitor receives the instructions to present information on the screen and when that information becomes visible. LCD manufactures typically report some type of rise- and fall-time, which is intended to reflect how long a pixel takes to switch from black-to-white and white-to-black. The display latency issue includes this rise and fall time, along with the time required for the image to work its way through the LCD circuitry.
When LCDs first became available, Psychology Software Tools recommended that they not be utilized, and that CRTs were the preferred monitor type, for the following reasons:
- The phosphor activation and decay on a CRT monitor are well-defined and widely understood, while the rise and fall times reported for LCD monitors were ill-defined and difficult to compare1. Vendors typically reported latencies for switching a pixel from black to white, which will show the fastest rise time. More recently, vendors have begun reporting gray-to gray specifications. However, there is no industry standard for this metric either, and the gray settings can be manipulated easily to suggest better results.
- CRTs outperformed early LCD monitors. CRT’s phosphor activation times are relatively constant and less than 2 ms, whereas the display latencies for early LCDs were often over 10 ms. However, our recent testing efforts have shown certain LCD monitors with acceptable display latencies. In general, we recommend that display latencies be less than on-half of the refresh rate.
In conclusion, our original recommendation against using LCD monitors no longer holds, albeit with the following cautions. First, you should always check the manufacturer’s specifications. Second, we continue to recommend that you test your LCD monitor with a third party test kit, such as the Black Box Toolkit, to determine if the monitor’s display latency is within your experimental requirements. While PST’s testing has identified LCD monitors that are capable of providing acceptable performance, we cannot provide a blanket endorsement of LCD monitors, either now or in the future. We continue to recommend testing in part because of potential technological changes that might be coming in LCD technology, such as painting the display in quadrants rather than line by line and various interlacing techniques. Such changes would almost certainly impact the timing of the displays.
Consequences of the refresh rate
There are three important consequences of the display having a fixed and non-varying refresh cycle. First, you cannot draw or remove a display at any time. Displays may be presented or removed only when the next refresh cycle comes along. For example, if the refresh cycle is 16 ms, displays may be changed at delays of 0, 16, 32… milliseconds from the time of the vertical blank that initiated the display sequence. The actual display of the stimuli that are half way down the screen will be halfway through the refresh cycle, or at 8, 24, and 40 ms.
Second, displays that are not a full refresh cycle in duration may possibly not be visible, and this varies depending on the position of the stimulus. For example, let us consider the same fixation-probe- mask experiment used above, but this time let us change the display specifications so that the fixation “+” is presented at the time of the first vertical blanking event, the probe word “CAT” is written 5 ms later, and finally the mask (“***”) is written 5 ms after the probe word. Given this stimulus timing, the “+” will actually never be presented to the participant, and instead the video monitor would present “CAT” at 8 ms and “***” at 24 ms (see Figure 3). This is because the “+” would be overwritten by” CAT” before it is every refreshed. The first refresh of the pixels in the center of the screen occur at time 8 ms, but the write of the probe “CAT” occurs at time 5 ms, which is 3 ms prior to the refresh of the pixels in that area of the screen.
Figure 3. Presenting displays for less than a refresh cycle produces partial displays. The arrows on the bottom show writing of the stimuli “+”, “CAT” and “***” to video memory at 0, 5, and 10 ms respectively. The participant sees “CAT” at 8 ms, and “***” at 24 ms and again at 40 ms. Note that the “+” is overwritten by “CAT” before it is ever displayed to the participant because it is not refreshed before it is overwritten.
The third consequence of the refresh cycle is that while you may specify a fixed display duration, if it is not an exact multiple of the refresh rate then the duration will vary up to one refresh cycle less than or greater than the desired time. For example, if the refresh rate is 60Hz (a refresh duration of 16 ms) and the specified display duration is 200 ms, the participant will see durations that are either 192 ms (12 refreshes) or 208 ms (13 refreshes). Since the requested duration (200 ms) is 50% of the distance between the onset times of the two refresh cycles (192 and 208), 50% of the refreshes will be at 192 ms and 50% at 208 ms.
A final concern with the refresh rate is that Windows and/or the video card manufacturers can mis- report the actual refresh rate being utilized. With respect to the operating system, Windows provides some type of interface to enable the user to select a refresh rate and screen resolution (for example, via the Display settings in the Control Panel). While this interface suggests that the user is setting the refresh rate, in actuality Windows treats this value as a request, and allows the display device driver to provide its best approximation to the request.
Further, with respect to the video card, the company that implemented the display driver code configured the software to display an acceptable image for business applications, but did not ensure that it was accurately reporting the refresh rate, or reporting back to Windows a close estimate of accuracy.
Auditory and Video Stimuli
As is the case with graphic files, time is required to read an audio file from the computer’s hard drive into memory. However, there are additional issues regarding timing when you are presenting auditory stimuli. The sound card, the sound device drivers, and the operating system all combine to determine the speed with which sound files can be played. When presenting sound files as stimuli, there is always a delay between when the audio card receives the command to play and the sound is actually played. This period is referred to as the sound latency. Sound latency varies between cards, but our testing also shows that it varies with the selected sound card driver, selected API (application programming interface), and operating system as well.