December, 2015
www.us-
tech.com Understanding Video Latency
A delay between what is seen on a display and what happens can be frustrating and problematic in many situations, such as a machine operator who would not want a delay in seeing the results of his movements. Similarly, those participating in a two-way video conference do not want delays between speaking and hearing a response. When choosing components for a video sys-
I
tem, it is important to understand how each piece affects the overall latency in the system. Achieving low latency, usually defined as less than 100ms, is most important for operating remote devices, video conferencing, streaming live events, and computer vision.
How Video Signals Are Displayed
Video is often described in terms
of frames per second (FPS). Two dif- ferent standard-definition analog video modes are used to describe this process. North and South America use the National Television System Committee (NTSC) definition, which is 29.97 FPS. Simply put, this means that 29.97 frames are displayed in any one-second period. The definition used in overseas locations is knows as Phase Alternating Line (PAL). It is usually 25 FPS, though some vari- ants exist. Therefore, when transmit- ting video signal, each frame takes 1/29.97 seconds (about 33.4ms) to be transferred for NTSC, and 1/25 sec- onds (40ms) for PAL. For NTSC, each frame is com-
prised of 480 lines of active video. In contrast, PAL includes 576 lines. Both NTSC and PAL video are inter- laced, meaning there are two fields for each frame. One field stores even lines while the other stores odd lines. A video camera does not capture both fields simultaneously; the camera shutter captures each field at twice the frame rate. For NTSC each field contains 240 lines of active video, while PAL fields contain 288. Suppose we have a video system
comprised of a video decoder (which translates an analog video to a digital signal), an image scaler, a video com- pression code, and a USB interface. The decoder must process the whole frame, including both fields, and write the interlaced image into mem- ory. The image scaler must wait until the entire frame is available before it can resize the image, though the operation can be done fairly quickly. The video compression codec must wait for the scaled image to be avail- able before compression, and this step is fairly lengthy. Once the frame has been compressed, the USB inter- face transfers the compressed video frame data to a host computer. The total time taken for the video frame to be acquired by the system and transferred to the host computer is the sum total of all these stages. Each stage works independently
of the others, and while one frame is being processed by the image scaler, the next is being captured by the video decoder. When choosing a video system,
operators typically specify how much latency they are willing to tolerate. Consider a scenario in which there is a requirement for a reduced-latency display (also called reduced-latency- preview) — meaning the operator needs to view the video at the same time as it is being recorded.
We didn't invent 3D AOI. We just perfected it.
n most cases, users want latency — the time it takes for a single frame of video to go from a camera to a display —to be as short as possible.
By Pete Eberlein, Sensoray Company, Inc. The device could bypass the image scaler and
video compression codec, and instead send the uncompressed frame directly to the host. Sending the
Is there a way to improve this? One way
The time taken for each step contributes to overall latency.
uncompressed frame over USB takes more time than a compressed frame as it requires a larger number of bytes. In this case, the video decoder requires 33ms per frame while the USB interface requires 20ms per uncompressed frame, for a total of 53ms.
11:34 AM
would be to use a video decoder that allows us to know when each field is complete by sending an interrupt. This would enable the system to quickly reassemble two fields into a frame on the host. We could then start sending one field early, while the video decoder is processing the next field. In this case the video decoder requires 33ms, or 17ms per field. The USB requires 20ms per uncompressed frame, or 10ms per field, so the whole frame would be transferred with a 10ms improvement. Some video decoders can be configured to inter-
rupt at a specific line interval. This would allow even lower latency by processing small chunks of lines instead of whole fields or frames. For example, if a programmed video decoder delivered an interrupt
Continued on page 55
Page 53
Inspect PCBs like a superhero.
With a new intuitive touchscreen interface packed with amazing features and extraordinary image quality.
Suppression (MRS) technology to let you see like never before. An easy-to-use software superhuman results, which is, metrology grade accuracy at production speed.
Watch the SQ3000 on YouTube at
http://bit.ly/1C9ouaM
www.cyberoptics.com Copyright © 2015. CyberOptics®
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88