![\"\"](\"docs\/digitizing_vhs\/digitizing_vhs.png\")
<\/p>\nThis article explains a setup and workflow for digitizing analog video (e.g. VHS, Beta, Video 2000, LaserDisc, …) using a Mac and digital camcorder \u2013 in high quality and with interlacing intact; optimized for archival. We will use a old-school digital camcorder (they are cheap!) to convert the analog signal to a high-quality digital “DV” stream and then record the DV stream on a Mac using a FireWire connection.<\/p>\n
<\/p>\n
Standard definition analog video is either<\/p>\n
Interlaced means that every full frame is split into two “fields”, one with the picture’s 288 (PAL; NTSC: 240) odd lines, and one with the 288 (240) even lines. These two fields can represent one moment in time: When combining them, they will give 25 (30) full 720×576 (720×480) frames. Or they can be recorded 1\/50 (1\/60) second apart, giving motion at a resolution of 50 (60) Hz, but at half the vertical resolution, and with the set of lines alternating every time.<\/p>\n
Interlaced video was natural for old CRT displays, but in order to show interlaced video on modern displays or encode them into modern video compression formats, they need to be converted into progressive format, i.e. deinterlaced.<\/p>\n
The two fields are always transmitted one after the other, and it is unclear whether every two fields should be combined (“comb filter”) for a 25 (30) Hz video or whether the 50 (60) fields should be vertically upscaled for a 50 (60) Hz video. Using the wrong method means either combing artifacts<\/a> or flickering.<\/p>\n The worst part is when the pairing of fields is inconsistent, like in this example of Futurama<\/a> or whenever two scenes are composited in the original SD version of Star Trek TNG.<\/p>\n So deinterlacing is hard, especially if you want to do it well. When digitizing analog video, it is best to keep the interlacing intact. When playing the file, VLC for instance will already do pretty good deinterlacing by default \u2013\u00a0and tomorrow’s VLC will certainly do a better job. And if you come back to the footage years later to share it or reuse parts in an HD video, you can use the best deinterlacer available then!<\/p>\n So all in all, if you want to archive<\/em> analog video, you should keep it interlaced. Many simple solutions won’t do this, but this solution does.<\/p>\n A full, uncompressed digital representation of a PAL signal is 50 times a second a 720×576 image at 24 bits per pixel, which is 25x720x576x24: almost 240 Mbits\/sec.<\/p>\n This implies “4:4:4”, meaning every 2×2 pixels have four brightness values (Y’) and four values each for the two color components (Cb and Cr). Most digital formats use chroma subsampling<\/a>, meaning that a 2×2 pixel grid has fewer chroma values. Since the human visual system is less sensitive to color than it is to brightness, 4:2:2 (half horizontal chroma resolution) is practically indistinguishable from 4:4:4.<\/p>\n The 1986 D-1 format<\/a> uses 4:2:2 chroma subsampling and thus reduces the data rate by a factor of 1.5 to about 160 MBits\/sec. Sony’s 1993 Digital Betacam<\/a> additionally uses lossy compression to reduce the data rate by a factor of 2.34:1 down to about 70 MBits\/sec. Both formats were meant for professional use.<\/p>\n The consumer format for digital SD camcorders is the 1994 DV<\/a> (“Digital Video”). It uses 4:2:0 chroma subsampling (one color value per 2×2) for PAL and 4:1:1 (one color value per 4×1) for NTSC, which reduces 4:4:4 data by a factor of 2. The resulting data is lossily DCT<\/a>-compressed by a factor of 5, leading to a data rate of 25 MBit\/sec.<\/p>\n DV embeds an uncompressed 48 kHz 16 bit stereo PCM audio stream, which adds 1.5 MBit\/sec (or alternatively, two 32 kHz 12 bit stereo streams with the same total bitrate).<\/p>\n Unlike today’s common compression methods (e.g. MPEG-2, H.264, H.265), DV compresses images individually, so in the stream, there are no dependencies between images. You can imagine DV as a stream of 64 KB JPEG images. This allows frame-exact editing and allows for simpler encoder and decoder hardware, but means that a higher data rate is required for the same quality. A 25 MBit\/sec DV stream is roughly equivalent to a 10 MBit\/sec MPEG-2 stream and a 3 MBit\/sec H.264 stream.<\/p>\n DV is an excellent match as an intermediate digital representation or even as an archival format for pretty much<\/a> all analog media, it even surpasses LaserDisc and matches DVD in luma and chroma resolution<\/a> (PAL values; NTSC are similar):<\/p>\nDigital Video and DV<\/h2>\n
![\"\"](\"docs\/digitizing_vhs\/4-6.jpg\")
<\/td>\n<\/tr>\n![\"\"](\"docs\/digitizing_vhs\/4-9.jpg\")
<\/td>\n<\/tr>\n![\"\"](\"docs\/digitizing_vhs\/fw_tb.jpg\")
<\/td>\n<\/tr>\n![\"\"](\"docs\/digitizing_vhs\/tb_tb.jpg\")
<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n