To Place an Order or get more info. Call Uniforce Sales and Engineering (510) 657 4000 www.uniforcesales.com Transitioning from NTSC (analog) to HD Digital Video Sheet 1
NTSC Analog Video NTSC video -color bar test pattern Sheet 2
SDI Digital Video At the SDI source At the end of a 100 meter cable Sheet 3
Making stills move There are two parts: Frame Rate (pictures per second) Set to give the illusion of smooth motion; beyond persistence of vision frequency. Rates above 16 images/second yield smooth motion 24 fps is used in film; 25 in the EC (PAL) and 30 in the USA (NTSC) Illumination Rate (most often 2x frame rate) Flicker fusion is the frequency that pulsing light looks steady Illumination rate is pushed high enough to achieve flicker fusion Film generally uses 48 Hz flicker rate, interlace TV scanning is 2x the frame rate. Sheet 4
Pixels Pixels are a Multifaceted Picture Element Number of Pixels is Only a part of the resolution story Shades of gray (steps, pixel depth) Few pixels Many shades Many pixels Few shades Sheet 5
Pixels Many pixels Many shades Specifying resolution Covers Pixel elements Shades of gray PER COLOR = number of colors Sheet 6
Pixel Depth; Color Depth; Colors 2 bits = 4 colors 4 bits = 16 colors 8 bits = 256 colors According to http://en.wikipedia.org/wiki/color, humans can distinguish up to 10 million colors 24 bits = 16 million colors Pictures from http://en.wikipedia.org/wiki/color_depth Sheet 7
Pixels & Bit Count & Data Rate Why is this all of this important? BIT Rate; Some Basics Bit rate pixel count X sampling & encoding method Sampling: The eye is more sensitive to intensity changes than color changes Subsampling is delivering fewer color samples than luma samples for a group of pixels 4:2:2 Subsampling = color at ½ luma rate Image quality indistinguishable from sampling both at the same frequency Sampling resolution is typically (TV) 10 bits per channel Channels are Y (luma), Cr (red component), Cb (blue component) Green derived from Y Cr & -Cb (similar to analog video) Sampling Frequency generally 74.xx MHz for HD Sheet 8
Quick Idea About Subsampling 4:2:2 subsampling causes two luma samples to share one pair (Cr and Cb) of color samples Color Sample Color Sample Active Video SDI Data Stream Cr 0-1 Y 0 Cb 0-1 Y 1 Cr 2-3 Y 2 Cb 2-3 Y 3 Sample Pair Sample Pair Sample Pair Sample Pair Pixel 0 Pixel 1 Pixel 2 Pixel 3 Graphic from Chrominance Subsampling in Digital Images, by Douglas Kerr Sheet 9
Pixels & Bit Count & Data Rate Key Points to Remember Each pixel = a Y (luma) sample Vertical blanking space adds lines; e.g. 45 in 1080 1080 lines, plus blanking = 1125 lines/frame Horizontal blanking space adds samples; e.g. 280 Y samples per line in 1080/60 1920 visible pixels + 280 Y H blanking samples = 2200 pixels/line Each pixel in 4:2:2 sampling is 20 bits deep 10 bits of luma (Y) and 1 of the color components (Cr or Cb) @ 10 bits = 20 bits Interlace Video delivers ½ the image in one field and the other half in a second field Frame rate = ½ field rate (e.g. 1080i/30) Progressive Video delivers a complete frame per scan Frame rate = field rate (e.g. 1080p/60) Frames may repeated at field rate (e.g. 1080p/30) Sheet 10
Pixels & Bit Count & Data Rate Calculating Bit Rates BIT RATE = Resolution x sample depth x Fields/Frame X Frame Rate For 1080p/60 = (1125 lines x 2200 pixels) x (20 bits/pixel) x 1 Field/Frame x 60 fps For 1080i/30 (60Hz field rate) = (49,500,000 bits/image) x 1 x 60 fps =2,970,000,000 bits/second =(1125 x 2200/2) x 20 bits/pixel x 60 fields/sec = 24,750,000/field x 60 = 1,485,000,000 bits/sec For 720p/60 ={(720+30) x (1280+370)} x 20 bits/pixel x 30 fps x 2 =24,750,000/image x 60 = 1,485,000,000 bits/sec Bit rate examples assume 4:2:2 subsampling IAW SMPT 259M Sheet 11
Electronics Two 32 Bit RISC Processors @ 100 MHz Custom 43K CE FPGA to substitutes image samples, sample x sample in real time Custom 25K CE FPGA design to manage ITS substitution engine, specify text and graphic overlays, colors and housekeeping functions 3 GHz Data I/O Pathway with equalizers & drivers One 8 bit Z80 Microprocessor running at 4 MHz Custom 320 CE Gate Array to hold text bit maps and manage overlay timing 20 MHz Video amp and coax driver Sheet 12
Transport How do you get raw video (SDI) from source to destination? Choices Analog NTSC any channel with a 6 MHz bandwidth will work SD-SDI Requires a channel capable of passing 143 MHz data rate HD-SDI Requires 1.5 GHz channel for 720p/1080i and 3 GHz for 1080p Direct Connect (copper) SMPTE Specs SDI be capable of operating to 100-200 meters of 75 Ω Coax (e.g Belden 1694A) These do require line equalizers and drivers; Reclocking is generally needed to properly decode Short runs can use Cat 6A/Cat 7 copper cable for short runs (10 m); CAT 5 and standard CAT 6 will not work Sheet 13
Transport Choices (cont.) Ethernet Fiber At 1G Ethernet, ED-SDI can work At 1 G Ethernet, 720p/1080i will not work 10G Fiber Only full duplex only 100G Fiber Only, full duplex, still evolving 10GBASE-ER single-mode fiber supports transport @ 10.3 Gbit/sec up to 30-40 Km Next level down -LR can support this rate up to 220 meters Radio Where would the band and bandwidth exist? Sheet 14
The Compression Beast Compression is a tool to reduce data rate Alternative to whole new infrastructures Typical Compression Ratios that maintain excellent image quality H.263 and MPEG-2 ; 30:1 MJPG 2000; 20:1 to 40:1 H.264/MEG-4 part 10; 50:1 Compression Issues Interframe prediction (MPEG) vs. image compression (M-JPG) MPEG is motion sensitive M-JPG can generate rings at the harsh image edges Trade off between image quality and frame rate/frame dropping Latency Sheet 15
The Compression Beast MPEG Coding Computationally Intensive More flexibility between image quality and frame rate tradeoff Sheet 16
The Compression Beast MJPEG Coding Less computationally intensive due to the lack of prediction Less bit efficient, will force tradeoff between frame rate sooner, image ringing Sheet 17
The Compression Beast Latency decode (sdi-image stream) + compression + xmit latency + buffer time + decompress + decode for display SDI Source Decode Compress xmit Display Decompress Buffer Receive Compression Many factors including image content, motion between frames, hardware speed Buffer Time Decompression requires a complete data set and enough buffered data to ensure every frame is reconstructed at the full expected frame rate 4-5 frames of data may be needed up to 20 depending compression parameters (MPEG) 83-300 ms seconds to complete a buffer @t 100 MB Ethernet @30:1 compression, 720p/1080i 74 ms to complete a buffer using MJPEG2000 @ 2-3 frames; @ 20:1 compression Sheet 18
Bit Rates SMPTE Standard Video Type Example Formats 259M SD-SDI 480i, 576i 344M 292M ED-SDI HD-SDI 480p, 576p 720p, 1080i 424M 3G-SDI 1080p Bit Rates (Mbits/s) 270 360 143 177 Bit Rates MJP2000 (Mbits/s) @10:1 27 36 14 18 Bit Rates MPEG-2 (Mbits/s) @ 30:1 9 12 4.8 5.9 Bit Rates MJP2000 (Mbits/s) @40:1 6.8 9 3.6 4.4 Bit Rates H.264 (Mbits/s) @50:1 5.4 7.2 2.9 3.5 1 Stream Video over Ethernet 10 Base T 540 54 18 13.5 10.8 100Base T 1485 1470 2970 2940 148 147 297 294 49.5 37 29.7 100Base T 99 74 59.4 1000 Base T Sheet 19
The Compression Beast Accuracy of time stamps at the destination (the method used in analog NTSC) is unpredictable do to wide variation in latency Transport mechanism Encoding/Decoding mechanism Amount of pre-image regen buffering Camera control more difficult due to image latency Conclusion? Time stamping must be at the source of the SDI digital video stream Transport necessities will not impact time stamping accuracy Manage transport bandwidth to minimize latency Sheet 20
The Compression Beast Degraded Image Quality Threat Degree of compression needed Video content Hardware CODEC speeds Transport bandwidth Degraded image quality issues Fine detail may be smeared or lost to macroblocks Overlay text may be smeared or unreadable Size of characters chosen for time stamping and other critical data at record time may not be appropriate displays at analysis and playback time Conclusion? Time stamp and store critical information in SDI metadata stream at the source Ancillary Packet Format (metadata) per SMPTE 291M and related specifications Survives compression losslessly Decoder can overlay at display time Sheet 21
HD Video Spec Checklist Use SDI video sources Must be SMPTE compliant Digital equivalent of the raw video Must preserve meta data Use recording devices that preserve metadata Metadata decoders can then place critical data on the video at playback Design your system such that Specify a system that stamps at a finite instance in the video (e.g. vertical sync) Time stamps and other time to image critical data is impressed into the SDI video stream and meta data at the source Eliminates any latency sources Genlock your video sources Sheet 22
HD Video Spec Checklist Avoid systems using standard SMPTE time stamp encoding SMPTE standard is accurate to the second, but only records frame number thereafter Use equipment that time stamps at a finite point in the SDI stream (e.g. vertical sync) Use equipment that captures time in fractions of a second e.g. 6980G-HD captures to 100 µs precision Use Equipment with interoperable metadata encoding STANAG 4609 is a possible method Is in use in several NATO and US programs Provides a non-proprietary format for encoding accurate time and other critical data Builds on and compliant with SMPTE 291M and related specifications Sheet 23
Comparing Analog to Digital Video Attribute Analog Video SDI Digital Video Raw Video Sync Complex AM, FM and phase modulated signal requiring 6 MHz bandwidth Pedestal and color burst sync areas scaled generally below the black level Serial encoded bit stream at bit rates from 140 Mbits/s to 3000 Mbits/sec A reserved bit pattern defined by SMPTE in the SDI stream Blanking Active Video Frame/Field Rate A predetermined voltage level in the video signal An AM signal with overlaid phase modulated color information RS 170 60Hz /30 Hz Field/Frame RS170A (NTSC) 59.94 (60/1.001) CCIR 50Hz /25Hz field/frame Progressive and Interlaced Fixed format data blocks before and after the active video data set A stream of video samples the number of which and format varies with resolution, sampling scheme and color depth 24 Many from 24.975 to 60 Hz and beyond Progressive and Interlaced Sheet 24
Comparing Analog to Digital Video Attribute Analog Video SDI Digital Video Visible Scan Lines Resolution/Line Color Sampling NTSC 480/frame PAL 576/frame This depends on the source and signal quality but ranges to the equivalent of 300 to 720 pixels Continuous time domain signal, intensity swings are limited by the available 1.5 MHz bandwidth 480 SD 576 (PAL) SD 720 HD 1080 HD SD 720 pixels HD 1280 and 1920 pixels YUV encoding samples intensity every pixel and color differently depending on the encoding chosen. 4:2:2 is most frequently used 25 Sheet 25
Interlaced (I) Progressive (P) Pixels & Bit Count, Data Rates, Resolutions & Specs SDTV HDTV Active lines per frame Total lines per frame Active Luma sample s per Line Luma Samples in Blanking Area Total Luma Samples (Pixels) Aspect Ratio Frame Rate (Hz) SMPTE SDI Bit Format Digitizing Bit Rate Specification (MBit/Sec) SDT V I 480 525 720 190 910 4:3 29.97 259M HDTV P 720 750 1280 370 1650 16:9 HDTV (PAL) 60 or 60/1.001 292M P 720 750 1280 700 1980 16:9 50 292M HDTV I 1080 1125 1920 280 2200 16:9 HDTV (PAL) 30 or 30/1.001 292M I 1080 1125 1920 720 2640 16:9 25 292M HDTV P 1080 1125 1920 28 0 2200 16:9 60 424M ITU-R BT.601 ITU-R BT.709 ITU-R BT.709 ITU-R BT.709 ITU-R BT.709 ITU-R BT.709 143.18 1485.0 1485.0 1485.0 1485.0 2970.0 Sheet 26
Uniforce Sales and Engineering is not just an exclusive Machine Vision Distributor for the leading brands, we Engineer Machine Vision solutions. To Place an order or get more information on ITS Products, Call Uniforce Sales today at (510) 657 4000 or www.uniforcesales.com White Paper Courtesy of ITS, Paul Hightower. Sheet 27