Video Intermediate
Introduction
The doors open in twenty minutes. The keynote laptop shows a perfect picture on its own screen and nothing on the LED wall. The media server's 4K clip flickers. The camera feed from the back of the hall sparkles every few seconds. None of the equipment is broken. Each fault is a signal arriving as something other than what the next device expects.
This book is about that gap. It goes inside the video signal: the interfaces that carry it and how far each goes, the handshakes that decide whether a picture is sent at all, the colour format, bit depth and range inside it, the arithmetic that says whether it fits a cable, what processing does to it and how long that takes, how it travels further, and how to measure it rather than trust a monitor.
It is built on published sources: the international standards for HD, UHD and HDR television, the HDCP specifications, the European recommendation on signal levels, a measurement guide from Tektronix, and the documentation of the makers whose equipment you will use. Where the book gives a number, it says where it comes from, and where it works one out, it shows the working.
How to Use This Book
Read Video Basics first. This book assumes it. Then read in order: the interfaces, the handshakes, what is inside the signal, processing, distance, measurement, and finally the method that uses all of it to find a fault.
Keep the arithmetic close
A handful of multiplications decide whether a signal fits a cable or a network. The book works each one through. Do them yourself for the formats you use most, and keep the answers with your kit.
A note on the words
A "source" sends a signal and a "sink" receives it. "Range" means narrow or full range, not distance. "Gb/s" and "Mb/s" are gigabits and megabits per second. The book uses "fiber", the industry's spelling, for optical cable.
Deeper Video Systems
Beyond the basics to working video knowledge.
From basics to systems
Video Basics introduced video — the camera, the image, basic signals. Video Intermediate develops deeper, working knowledge of video systems: how video signals really work (resolution, frame rate, colour), the formats and standards that carry them, the processing that converts and shapes them, and the signal paths that move them through a system. The goal is real working understanding of video — enough to work with professional video systems and signals confidently, and to build toward video engineering.
This matters because professional video work involves dealing with many signal types, formats, resolutions, and conversions, and understanding video systems deeply is what lets you set them up, connect them, and troubleshoot them. The basics get an image; the systems knowledge lets you work professionally with video.
Video Basics teaches what video is, where pictures come from, the signal types and connectors you will meet, what the boxes in the middle are for, the displays at the end, and how to trace a chain. It introduces EDID, HDCP, full and narrow range and video over a network well enough to recognise them. This book assumes all of that and goes inside each one.
Video Intermediate is the book for the person who makes the picture actually arrive. It covers the professional and consumer interfaces in enough depth to choose cables and predict distance; the two silent conversations, EDID and HDCP, behind most black screens; what is inside a signal, from colour format to bit depth to range, and the arithmetic that says whether it will fit down a cable; what conversion and processing do to a picture and to its timing; how signals travel further, over fiber, HDBaseT and networks; how to measure a signal rather than judge it by eye; and how to find a fault by method.
Two things are deliberately left to other books. Keeping every source locked to one timing reference, genlock, is Video Engineering's subject, along with matching cameras. And the large switchers and routers of big productions are Advanced Switching and Signal Routing's.
Video as signals and systems
At the intermediate level, video is understood as signals moving through systems. A video signal carries the image (with its resolution, frame rate, and colour), in some format (SDI, HDMI, IP, etc.), through a system (sources, processing, routing, distribution, displays). Understanding video as signals and systems — what the signal is, how it's carried, and how it moves — is the framework for everything in professional video. This systems view (building on signal flow from AV Fundamentals) is what intermediate video develops.
This framework lets you reason about any video setup: what's the signal, what format, what path, where might problems be. It's the mental model of professional video work. The rest of this book fills in the signals, formats, processing, and paths.
Practice it
A colleague says the picture problems on last week's show were "just the cables". The faults were a laptop that fell back to a low resolution, a grey, flat picture on the LED wall, and sparkles on a long camera run.
- You have worked through Video Basics.
- You can trace a video chain from source to display.
- You know what EDID, HDCP and full or narrow range are, if not yet how they work.
SDI and HDMI in Depth
The two interfaces that carry most event video were built for different worlds. SDI was made for broadcast: long cables, locking connectors, no copy protection. HDMI was made for living rooms: short cables, a conversation between devices, and protection built in.
The SDI family
SDI has grown by doubling. Standard definition ran at 270 megabits a second. HD-SDI, SMPTE ST 292, runs at 1.485 gigabits. 3G-SDI, SMPTE ST 424 and 425, runs at 2.97 for 1080p at 50 and 60 frames. 6G and 12G, SMPTE ST 2081 and 2082, run at 5.94 and 11.88 for UHD. Each rate also has a version divided by 1.001 for 59.94-frame countries, so AJA's 12G distribution amplifier lists 1.483, 2.967, 5.934 and 11.868 as well.
3G comes in two kinds. Level A maps the picture directly. Level B carries two HD-SDI links packed inside one 3G stream, the old dual-link arrangement. Most modern equipment handles both, but an older device that accepts only one shows nothing at all from the other, so check which a source sends when a 3G picture refuses to appear.
UHD can travel as one 12G signal or be split across several cables. ITU-R BT.2077 lists 4K at 60 frames in 10-bit 4:2:2 as one 12G link or two 6G links; 4:4:4 needs two 12G links or four 6G. Quad-link 3G carries it on four 3G cables. A quad-link picture can be divided in two ways: four quadrants, called square division, or two-sample interleave, where each cable carries a full-frame picture made from every other pair of pixels. Sender and receiver must agree, and AJA's FS4 manual notes that square-division video carries no payload identifier to say which it is.
How far SDI goes
Higher rates go less far on the same cable, because coax loses more at higher frequencies. Tektronix's guide gives standard definition at about 300 metres and HD-SDI at about 100. ITU-R BT.2077 specifies UHD links over coax at under 100 metres for 6G, under 70 for 12G and under 30 for 24G.
Real equipment publishes its own figures, which depend on how well its receiver equalises the cable. AJA rates its 12GDA amplifier on Belden 1694A at 70 metres for 12G, 120 for 6G, 190 for 3G, 195 for HD-SDI and 405 for standard definition. Belden's own engineers point out that published distances are calculated, cautious figures. Plan to the equipment maker's numbers for the cable you have, and treat anything longer as a test, not a promise.
SDI fails abruptly. The receiver recovers a perfect picture until the signal is too weak, and then the picture breaks up, sparkles or vanishes, with very little warning in between. Tektronix calls it the digital cliff, and notes that occasional CRC errors, one a minute or one a second, are the sign of a link standing at its edge.
| Signal | Data rate | Guide or standard distance | Source | AJA 12GDA on Belden 1694A |
|---|---|---|---|---|
| SD-SDI | 270 Mb/s | about 300 m | Tektronix digital video guide | 405 m |
| HD-SDI | 1.485 Gb/s | about 100 m | Tektronix digital video guide | 195 m |
| 3G-SDI | 2.97 Gb/s | none given | — | 190 m |
| 6G-SDI | 5.94 Gb/s | under 100 m | ITU-R BT.2077-3, Table 1 | 120 m |
| 12G-SDI | 11.88 Gb/s | under 70 m | ITU-R BT.2077-3, Table 1 | 70 m |
The HDMI versions
HDMI versions are mostly about bandwidth. HDMI 1.4 carries up to 10.2 gigabits a second, enough for 4K at 30 frames. HDMI 2.0 carries 18, enough for 4K at 60. HDMI 2.1 carries 48, enough for 8K at 60 or 4K at 120, and the HDMI Licensing Administrator now describes a 2.2 with up to 96.
The jump to 2.1 changed how the signal travels. Earlier versions used TMDS: three data pairs and a separate clock, with ten bits sent for every eight. HDMI 2.1's fixed rate link, FRL, uses all four pairs for data at up to 12 gigabits each, embeds the clock, and codes 16 bits as 18, which wastes less of the cable.
Cables are certified by speed, not by version number. A High Speed cable is rated to 10.2 gigabits, a Premium High Speed to 18 and an Ultra High Speed to 48. A 4K60 HDR signal down an old High Speed cable may work, flicker, or drop out every few seconds, and the cure is the right cable, not a different setting.
DisplayPort and USB-C
Laptops increasingly output DisplayPort, often through a USB-C port. DisplayPort 1.4's HBR3 carries 8.1 gigabits on each of four lanes, 32.4 in total. DisplayPort 2.1 reaches 20 gigabits a lane, 80 in total, on DP80-rated cables, and uses a more efficient code with about 3 percent overhead instead of 20.
Through USB-C, DisplayPort travels in alternate mode, and VESA notes it may use only two of the four high-speed lanes so the port can still carry USB data. A USB-C hub that also runs a keyboard and a drive may therefore give the display half the bandwidth it expected. When a USB-C laptop will not send 4K at 60, a direct USB-C to HDMI or DisplayPort adapter is the first thing to try.
Practice it
You need to run 4K at 50 frames from a camera position 85 metres from the switcher, over coax.
- You can list the SDI rates and roughly how far each runs on coax.
- You know Level A from Level B, and square division from two-sample interleave.
- You know which HDMI cable rating each format needs.
- You know why a USB-C hub can halve a display's bandwidth.
EDID and HDCP
Most black screens at a show are not broken equipment. They are two quiet conversations between devices, one about what the display can show and one about whether it may be shown at all.
What EDID says
EDID, extended display identification data, is a small record a display keeps to describe itself. Extron's guide to EDID sets out its layout. The base block is 128 bytes: a header, the maker and model, the EDID version, basic parameters, colour characteristics, lists of supported timings, and four 18-byte detailed timing descriptors, the first of which is the display's preferred timing. Extension blocks of 128 bytes each follow, most often the CTA-861 block that lists video formats, colour formats and audio.
The source reads the EDID over a separate low-speed connection in the cable, the display data channel, which the source powers with 5 volts so it works even with the display switched off. A hot plug detect line tells the source a display has been connected, and the source reads the EDID and chooses an output.
When EDID goes wrong
Extron lists the symptoms every technician meets. No image at all. A computer that falls back to 640 by 480 or 1024 by 768 because it could not read a sensible EDID. And a mismatch when one source feeds several displays, because the source reads only one EDID and picks a format that suits that one.
Switchers, scalers, extenders and splitters all sit between the source and the display, and each either passes the display's EDID back or presents its own. That is why a laptop that worked a minute ago chooses a different resolution after a cable is reseated: the conversation happened again with a different partner.
The professional cure is to manage EDID deliberately. Most switchers, scalers and converters can present a fixed EDID that says exactly what format to send, AJA's HA5-12G among them. Set every input that faces a laptop or media server to the show's format, and the source will send that format every time, whatever is plugged in behind.
What HDCP does
HDCP, high-bandwidth digital content protection, encrypts a picture between devices so that protected content can only reach approved displays. Before sending, the source checks the device it is connected to. A repeater, such as a switcher or splitter, must in turn check everything connected to it and report back how many devices there are and how deep the chain goes.
The specifications set hard limits. HDCP 1.4 allows up to seven levels of repeaters and 128 devices in total. HDCP 2.3 allows four levels and 32 devices. Exceed them and the source stops sending. On a show with splitters feeding splitters feeding displays, the limit is easily reached.
HDCP 2.3 also defines two types of content. Type 0 may go to any HDCP device. Type 1 must not be sent to devices that support only HDCP 1.x, 2.0 or 2.1, so a chain that works for one film can go black for another.
HDCP and SDI
SDI has no HDCP. AJA's HA5-12G manual says it plainly: HDCP decryption is not supported, because by definition SDI outputs are unencrypted. A protected HDMI source therefore cannot be converted to SDI by a legitimate converter, and it cannot enter an SDI switcher.
In practice the protected sources at events are streaming services, some media players, and laptops playing protected video. Ask about them before the show. Content made for the event is almost never protected; the problem is the clip a presenter plays from a streaming service, or a laptop that protects its whole output because one window is playing protected video. Ask for the file instead, played from a show machine.
How to do it
- Decide the show's format for every computer and media input.
- Set each switcher, scaler or converter input facing a source to present a fixed EDID in that format.
- Connect each source and confirm it now outputs the show format.
- Disconnect and reconnect each source once, and confirm the format holds.
- Ask every presenter whether they will play protected content, and plan an unprotected copy.
What goes wrong
| Symptom | Likely cause | What to do |
|---|---|---|
| A laptop falls back to 1024 by 768 | It could not read a usable EDID through the chain | Present a fixed EDID from the input it connects to |
| Resolution changes after a cable is reseated | EDID was renegotiated with a different device | Manage EDID at the switcher instead of passing it through |
| Picture on a hotel TV, black through the switcher | The source is sending HDCP-protected content | Use an unprotected copy on a show machine |
| The last displays on a splitter chain go black | The HDCP device or repeater limit was exceeded | Reduce the cascade depth, or remove HDCP from the chain |
Practice it
A conference has a lectern laptop, a media server and a presenter's own laptop, all into a scaler, then to an LED processor, a confidence monitor and a recorder.
- You can say what EDID contains and how a source reads it.
- You present a fixed EDID to every computer input.
- You know the HDCP device and depth limits for 1.4 and 2.3.
- You know protected sources cannot go to SDI, and plan around them.
Video Signals
Resolution, frame rate, and colour.
Resolution and frame rate
Two fundamental properties of a video signal are resolution and frame rate. Resolution is the image's detail — the number of pixels (e.g., 1920×1080 'HD', 3840×2160 '4K/UHD') — more pixels meaning more detail. Frame rate is how many frames (images) per second (e.g., 24, 25, 30, 50, 60 fps), which affects motion smoothness. Together with whether it's progressive (p) or interlaced (i), these define the video format's image (e.g., 1080p60).
Professional video work requires knowing and matching these (a system's sources, processing, and displays must agree on or convert between resolutions and frame rates), and choosing them appropriately. Mismatched resolution or frame rate causes problems (no image, conversion issues, judder).
Colour and signal specs
Video signals also have colour and other specifications. Colour involves colour space/gamut (the range of colours), colour sampling (e.g., 4:2:2, 4:4:4 — how much colour detail), bit depth (the precision of colour/brightness, e.g., 8-bit, 10-bit), and standards (e.g., Rec.709, HDR) — all affecting how colour and tone are represented. Plus other specs (signal levels, sync, etc.). At the intermediate level, you become aware of these signal specifications and why they matter (correct, matched specs give correct images).
You don't need to master every spec at once, but understanding that video signals have these properties — and that they must be correct and matched through a system — is key intermediate knowledge. Mismatched or wrong specs cause colour and image problems.
RGB and YCbCr
A picture can travel as three colour signals, red, green and blue, or as one brightness signal and two colour-difference signals, written Y, Cb and Cr. Computers and graphics usually work in RGB. Broadcast video works in YCbCr, because the eye is far more sensitive to detail in brightness than in colour, so the colour signals can be sent at lower resolution and save bandwidth.
ITU-R BT.2100 defines how far that saving goes. In 4:4:4 every pixel keeps its own colour. In 4:2:2 the colour is sampled half as often across each line. In 4:2:0 it is halved across and down. Camera Theory explains what each costs a key; for a signal, the point is that 4:2:2 carries two thirds of the data of 4:4:4, and 4:2:0 carries half.
Every device in a chain has to agree on which one it is receiving. When a sink treats YCbCr as RGB, or the other way round, the picture arrives with strong false colours, usually pink, green or purple, while brightness looks roughly right. It is one of the most recognisable faults in video, and it is always a setting.
Narrow range and full range
Video does not use every number available for picture. ITU-R BT.709 puts black at code 16 and nominal white at 235 in 8-bit video, and at 64 and 940 in 10-bit, with colour-difference signals between 16 and 240, or 64 and 960. The space above and below holds super-whites and sub-blacks. BT.2100 calls this the narrow range and says it is the default; full range, 0 to 1023 in 10-bit, is not to be used for programme exchange unless everybody involved agrees. Computer graphics, though, usually run full range, 0 to 255 in 8-bit.
The trouble starts when one device sends one and the next assumes the other. Work through the numbers. A narrow-range signal shown as full range puts its black, code 16, at about 6 percent grey, and its white, code 235, at about 92 percent. The picture loses its true black and its brightest white, and looks grey and flat. A full-range signal shown as narrow range does the opposite: everything below 16 crushes to black and everything above 235 clips to white, so dark detail and bright detail vanish.
The cure is a setting, on the source, a converter or the display. AJA's Hi5-12G converter, for example, lets an RGB input be read as SMPTE or full range, and its output be set to automatic, SMPTE or full. Find the stage where the range assumed stops matching the range sent, and set it there.
Timing, pixel clocks and bandwidth
A video signal carries more than the picture. Each line has extra samples, and each frame extra lines, that carry no image and leave room for timing, audio and data. Tektronix's guide to digital video gives 1080-line video at 60 frames as 2,200 samples per line and 1,125 lines per frame, of which 1,920 by 1,080 are picture. Multiply them by 60 frames and you get the pixel clock: 148.5 million samples a second, 148.5 megahertz.
That number lets you work out whether a signal fits a cable. At 1080p60 in 8-bit RGB, each of 148.5 million pixels carries 24 bits, about 3.6 gigabits a second. HDMI's TMDS coding sends ten bits for every eight, so the cable must carry about 4.5. HDMI 1.4 handles 10.2 gigabits, so it fits easily.
Now try 4K at 60 frames. Rohde and Schwarz's HDMI reference gives its pixel clock as 594 megahertz. In 8-bit RGB that is about 14.3 gigabits of picture, 17.8 on the wire: just under HDMI 2.0's 18.
In 10-bit it is 594 million times 30 bits, about 17.8 gigabits of picture and 22.3 on the wire, which does not fit. That is why 4K60 in 10-bit on HDMI 2.0 has to drop to 4:2:2 or 4:2:0, and why full-quality 4K60 HDR needs HDMI 2.1.
SDI is worked the same way. HD-SDI carries a 74.25 megahertz luma sample stream and a colour stream of the same rate, 10 bits each: 74.25 million times 2 times 10 is 1.485 gigabits, the HD-SDI rate. Double the clock for 1080p60 and you get 2.97 gigabits, the 3G rate. Four of those make about 12 gigabits, 12G.
HDR signalling
An HDR picture is only shown correctly if the display knows it is HDR and which curve it uses. That information travels beside the picture. On HDMI it is carried in InfoFrames, small packets that describe the transfer curve, the colour space and, for PQ, static metadata: the mastering display's colour and brightness from SMPTE ST 2086, and the content's maximum light levels, MaxCLL and MaxFALL. On SDI the curve and colour space are flagged in the payload identifier, the VPID.
AJA's Hi5-12G manual makes two points worth keeping. Its HDMI HDR InfoFrames are generated only when the signal is marked PQ or HLG. And the metadata does not change the picture data at all; it only tells the display how to show it. So a correct HDR signal with missing or wrong metadata reaches the screen intact and is displayed wrongly, usually flat and washed out. Check the flags before blaming the picture.
Two fundamental properties of a video signal are resolution (the image's detail — pixel count like 1920×1080 HD or 3840×2160 4K, more pixels = more detail) and frame rate (frames per second like 24/30/60, affecting motion smoothness), which with progressive/interlaced define the format (e.g., 1080p60) — they must be known and matched across sources, processing, and displays (mismatches cause no image, conversion issues, or judder). Signals also carry colour and other specs — colour space/gamut, sampling (4:2:2, 4:4:4), bit depth (8/10-bit), standards (Rec.709, HDR), signal levels, sync — which must be correct and matched through a system to give correct images; intermediate knowledge means being aware of these and why they matter.
Practice it
A media server sends 4K at 60 frames in 10-bit 4:4:4 into a switcher input labelled HDMI 2.0.
- You can say why YCbCr and 4:2:2 save bandwidth.
- You can name the symptoms of each range mismatch.
- You can work out whether a format fits a cable from its pixel clock.
- You know HDR metadata labels a picture rather than changing it.
Formats, Standards & Processing
How signals are carried and shaped.
Formats and standards
Video signals are carried in formats over various standards/interfaces. SDI (Serial Digital Interface, over BNC/coax) is the professional broadcast/production video standard. HDMI (and DisplayPort) are common in consumer/pro AV. IP/AV-over-IP (video over networks — NDI, SMPTE 2110, and others) is increasingly important. And codecs/compression (how video is encoded for storage/transmission, e.g., for files and streaming) matter for recorded and streamed video.
Each has its characteristics, uses, and connectors, and professional video work involves dealing with and converting between them. Understanding the formats and standards lets you connect and route video correctly.
Video processing
Video systems process signals — and understanding processing is key intermediate knowledge. Conversion changes a signal between formats (e.g., SDI to HDMI, or one resolution/frame rate to another). Scaling resizes the image (up or down) to fit a different resolution. Sync/timing keeps signals aligned (genlock and timing, so signals can be switched and combined cleanly). And other processing (colour, frame rate conversion, etc.) shapes signals. Processing is how video systems make different signals work together.
This processing is essential because real systems mix signal types, formats, and resolutions that must be made compatible (converted, scaled, synced) to work together (e.g., to switch between them, or show them on a display). Understanding what processing does (and that it's needed) lets you build and troubleshoot systems.
Scaling, conversion and frame-rate conversion
A scaler changes a picture's size: up, down, or across from one aspect ratio to another. It has to invent pixels when it enlarges and discard them when it shrinks, so every scale costs a little sharpness. Scaling once, to the final format, is better than scaling several times along a chain.
Aspect ratio is a decision, not a setting to leave alone. A 4:3 picture in a 16:9 frame can be shown with black bars at the sides, stretched to fill, which makes people look wide, or cropped to fill, which loses the top and bottom. Know which the show wants.
Frame-rate conversion is the hardest job a processor does. Going from 59.94 frames to 50 means showing five frames for every six, and simple converters do it by dropping or repeating frames, which shows as a regular stutter on anything moving. Better converters build new frames from motion, at the cost of more delay and the occasional strange edge. Avoid frame-rate conversion wherever the show allows, by running everything at one rate.
Interlace and deinterlacing
Interlaced video, the i in 1080i, sends each frame as two fields, one of the odd lines and one of the even, captured a moment apart. Every flat panel and projector is progressive, so interlaced video has to be deinterlaced before it is shown. A poor or missing deinterlacer leaves combing, jagged horizontal teeth on the edges of anything moving. If combing appears, find which device is receiving interlaced video and which should be deinterlacing it.
Frame synchronisers, and the delay they add
A frame synchroniser stores the picture and sends it out again locked to the system's timing, which is how an unsynchronised source, such as a laptop or a wireless camera, can enter a switcher cleanly. Storing a frame takes time.
AJA's FS4 manual gives typical figures for one such box: about two frames of delay at 30 frames a second and below, about four at 50 and 60, and three or five when it is also converting between HD and UHD. Up to 20 more frames can be added on purpose. At 60 frames, four frames is about 67 milliseconds.
Not every box delays by frames. AJA lists its 12G audio embedder, which has no frame store, at about 3 microseconds. The rule is that anything that stores a frame, which includes frame syncs, scalers and most converters, adds at least a frame, and the delays add up along a chain. On IMAG, where the room hears the voice directly, that total is what the audience sees as late lips.
Video signals are carried in formats over standards/interfaces: SDI (over BNC/coax — the professional broadcast/production standard), HDMI/DisplayPort (consumer/pro AV), IP/AV-over-IP (video over networks — NDI, SMPTE 2110, increasingly important), and codecs/compression (encoding for files and streaming) — each with its characteristics, uses, and connectors, and professional work involves converting between them. Video systems also process signals: conversion (between formats), scaling (resizing the image to a different resolution), sync/timing (genlock keeping signals aligned so they switch and combine cleanly), and other processing — essential because real systems mix signal types, formats, and resolutions that must be made compatible to work together.
Practice it
A wireless camera at 59.94 frames feeds a frame synchroniser, then a scaler, then an LED processor, on a show running at 50 frames.
- You scale once, to the final format.
- You choose how a picture of another aspect ratio is shown.
- You avoid frame-rate conversion by running one rate.
- You can recognise combing and find the missing deinterlacer.
- You add up the delay of every frame store in an IMAG path.
Moving Signals Further
Every interface has a distance beyond which it stops. Beyond that, a signal is carried by something else: light in a fiber, a category cable, or a network.
Fiber
Fiber carries SDI as light. ITU-R BT.2077 gives single-mode fiber links for UHD at up to 2 or 4 kilometres depending on the part of the standard, and AJA rates its FiDO fiber converters at up to 10 kilometres on single-mode fiber for every rate up to 12G. Fiber also breaks the electrical connection between two places, so a ground loop between a truck and a stage cannot travel down it.
Fiber is unforgiving of dirt. A single speck on a connector end face can take a link down, as Camera Basics explains for camera cables. Clean and inspect every fiber connection before making it.
HDBaseT
HDBaseT carries uncompressed video over category cable. The HDBaseT Alliance gives its reach as up to 100 metres, and its 5Play feature set as video and audio, USB, Ethernet, control and power on the one cable. The versions matter. The 2013 version reaches 100 metres with 4K at 30 frames in 4:4:4 or 4K at 60 in 4:2:0 on Cat6. The 2019 version doubles the data rate to carry 4K at 60 in 4:4:4 to 100 metres, on recommended shielded Cat6a.
HDBaseT extenders are HDMI devices at each end, so EDID and HDCP pass through them, or fail through them, as through any repeater. And the cable quality matters: unshielded or poorly terminated cable, or a run through patch panels, can cut the real distance well below the rated one.
Video over a network
Video Basics introduces video over a network. The working choice is between compressing the picture to fit an ordinary network and building a network fast enough to carry it uncompressed. Uncompressed 1080p60 in 10-bit 4:2:2 is 1,920 by 1,080 pixels, 60 times a second, at 20 bits each: about 2.5 gigabits a second, more than a gigabit network carries.
NDI compresses. NDI's own documentation gives its High Bandwidth format at about 132 megabits a second for 1080p60 and 250 for 2160p60, in 8-bit 4:2:2, and its HX formats far lower, about 16 megabits for 1080p60 in HX2 and 62 in HX3, using H.264. NDI states a technical latency of 16 video lines, with most implementations in practice delaying about one field, and notes the network itself is not deterministic, so delay can vary.
SRT is a protocol for sending compressed video across the internet reliably. Its open-source project, led by Haivision, describes it as sub-second latency transport that recovers lost packets by asking for them again, encrypts with AES, holds a constant end-to-end delay, and carries any codec. It is how many remote contributions and stream feeds travel.
SDVoE builds a 10-gigabit network instead. The SDVoE Alliance describes uncompressed 4K at 60 frames in 4:4:4 over 10-gigabit Ethernet. It suits installed systems where many sources reach many displays.
SMPTE ST 2110 is the broadcast standard for uncompressed video on a network. SMPTE's roadmap describes it as separate streams for each essence, all locked to a common clock: ST 2110-20 for video, -30 for audio, built on AES67, and -40 for ancillary data, with the clock distributed by precision time protocol under ST 2059. The older ST 2022-6 carried a whole SDI signal as one stream; 2110 exists because the industry wanted to switch video, audio and data separately.
| Transport | What it carries | Bandwidth | Source |
|---|---|---|---|
| NDI High Bandwidth | Compressed, 8-bit 4:2:2 | about 132 Mb/s at 1080p60 | NDI documentation |
| NDI HX2 | H.264, 8-bit 4:2:0 | about 16 Mb/s at 1080p60 | NDI documentation |
| SRT | Any compressed stream, over the internet | set by the stream | SRT project (Haivision) |
| SDVoE | Uncompressed 4K60 4:4:4 | 10 Gb Ethernet | SDVoE Alliance |
| SMPTE ST 2110 | Uncompressed essences, separately | about 2.5 Gb/s per 1080p60 10-bit 4:2:2 stream | SMPTE OV 2110-0, worked |
What goes wrong
| Symptom | Likely cause | What to do |
|---|---|---|
| An HDBaseT link works on the bench and fails on site | Cable quality or patch panels cut the real distance | Use the recommended cable in one run, and test before the show |
| An NDI source arrives late or varies | Compression and a non-deterministic network | Keep NDI out of IMAG paths, or measure its delay first |
| A fiber link that worked yesterday is down | Dirt on a connector | Clean and inspect both ends before reconnecting |
Practice it
A 1080p60 camera is 400 metres from the control room, and a streaming team wants two cameras as NDI on the venue's gigabit network.
- You choose fiber when distance or ground loops demand it.
- You know which HDBaseT version carries which 4K format.
- You can work out an uncompressed bandwidth.
- You know what NDI, SRT, SDVoE and ST 2110 each do, and their costs.
Measuring the Signal
A monitor shows what a signal looks like on that monitor. A waveform monitor and vectorscope show what the signal actually is.
Colour bars and PLUGE
Colour bars are the reference everyone shares. SMPTE RP 219 describes the HD multiformat pattern: 75 percent bars in the middle, 100 percent cyan and blue below, a ramp from black to white for finding faulty bits, and a black row containing the PLUGE, a set of patches just below, at and just above black.
The PLUGE is for setting a monitor's black. Its steps are at minus 2, 0, plus 2, 0 and plus 4 percent. The standard's method: adjust brightness until the plus 2 and plus 4 percent steps are visible and the minus 2 percent step is not. A monitor set that way shows black as black, and anyone judging a picture on it is judging the signal, not the monitor.
Waveform and vectorscope
A waveform monitor plots brightness from bottom to top across the picture. Video Basics and Camera Basics use it to expose; here it is a test instrument. Colour bars on a waveform form a staircase of known heights, and a wrong height says a level is wrong.
A vectorscope plots colour. Tektronix's guide describes it: Cr, the red-difference signal, vertically, Cb horizontally, and a set of target boxes where each colour bar should land, for 75 or 100 percent bars. Dots outside their boxes mean a colour gain or phase error; dots too high in the upper half mean too much Cr. The scope's setting must match the bars being sent.
Tektronix also explains gamut. A YCbCr signal can be legal, within its own limits, and still describe colours that cannot exist in RGB, which become illegal the moment the signal is converted. Only about a quarter of possible YCbCr values map to valid RGB. EBU R 103 sets the limits engineers check against: RGB components between minus 5 and plus 105 percent, ignoring errors smaller than 1 percent of the picture.
Checking SDI itself
An SDI analyser can measure the electrical signal, not just the picture. Tektronix gives SDI's launch level as 800 millivolts and its allowed jitter as 0.2 of a bit period, which at HD rates is about 135 picoseconds. An eye diagram overlays thousands of bits: an open eye is a healthy signal, a closing eye is a cable or connector heading for the cliff.
Most working technicians will not carry an eye-pattern analyser. They will see the result: CRC errors counted by a monitor, converter or router, which say a link is losing data before the picture visibly fails. A link that logs errors before the show will fail during it.
Multiviewers as a check
The multiviewer in the control room shows every source at once, and most also show each input's format, audio meters, tally, and alarms for a missing or frozen picture. Used deliberately, it is a monitoring tool. Check the format reported on every input matches the plan, and set its alarms so a frozen or lost source is seen before the director cuts to it.
How to do it
- Send colour bars with PLUGE from the start of the path.
- At the end, check black on the PLUGE and the bars on a waveform and vectorscope.
- Read the format each device reports along the way.
- Check for CRC errors on every SDI link.
- Then repeat with the real source.
Practice it
Colour bars sent through a converter look right on a monitor, but the vectorscope shows every dot a little outside its box, rotated slightly clockwise.
- You can set a monitor's black with PLUGE.
- You can read colour bars on a waveform and vectorscope.
- You know legal from valid, and what gamut errors are.
- You watch CRC errors as the warning before an SDI link fails.
Signal Paths & Troubleshooting
Trace the path and solve problems.
The video signal path
At the intermediate level, you understand and trace video signal paths — the route a video signal takes through a system, from source through any conversion/processing, routing/switching, and distribution, to the display or recorder. Understanding the signal path (building on signal flow) lets you see how a video system works: where the signal comes from, what happens to it, and where it goes. Tracing the video signal path is a key skill — it's how you understand, set up, and troubleshoot a video system.
This means being able to follow any video signal through its system (this camera → this converter → this switcher → this scaler → this display), understanding each stage. A clear grasp of the signal path is the foundation for working with and fixing video systems.
Troubleshooting video
With signals, formats, processing, and paths understood, you can troubleshoot video problems — diagnosing and fixing the common issues (no image, wrong/distorted image, colour problems, signal dropouts, format mismatches). The intermediate approach is systematic (from Troubleshooting Methodology): trace the signal path, check each stage (is the source outputting, is the format right, is the conversion/scaling correct, is it routed right, is the display set right), and find where the problem is. Understanding the system lets you troubleshoot it logically rather than guessing.
Most video problems come down to signal-path or format/spec issues (a mismatch, a wrong setting, a bad connection, a failed device), found by methodically checking the path. The professional uses their systems knowledge to locate and fix the problem.
A method, not a guess
Most video faults are found in minutes by a method and in hours by guessing. The method is the same on every system. Know what the signal should be at every stage: its interface, resolution, frame rate, colour format and range. Then find the first stage where it stops being that.
- Start from a known-good source. A test pattern from a signal generator, a converter or the switcher proves the path without depending on somebody's laptop.
- Split the path in half. Check the signal in the middle of the chain; whichever half is faulty is the half you split next. Ten devices take at most four checks.
- Read the format at each stage rather than assuming it. Most switchers, scalers, converters and analysers will tell you what they are receiving. A fault is usually the stage where what arrives differs from what was expected.
- Change one thing at a time, and write each change down. Two changes at once can fix the fault and hide the cause, which comes back during the show.
- Substitute rather than repair during a show. A known-good cable, a spare converter or a different input gets the picture back; finding out why can wait for the break.
Faults by what you see
Each kind of fault has a look, and the look points to the stage. The chapters of this book explain each cause; this table is the short version to keep by the switcher.
| What you see | Most likely cause | Where to look |
|---|---|---|
| No signal at all | Source not outputting, wrong port, failed cable, or a format the input will not accept | Source output settings, then a known-good cable, then the input's accepted formats |
| Picture on preview, black on the output | HDCP, or a format the next device will not take | Whether the source is protected; the output format |
| Sparkles, snow or dropouts on SDI | Cable too long for the data rate, a damaged cable or connector, or cable not rated for the rate | Cable length and type against the rate |
| HDMI picture blinks or drops out | Bandwidth beyond the cable, or EDID and HDCP renegotiating | Cable rating, format, and whether something was replugged |
| Blacks grey and the picture flat | A narrow-range signal shown as full range | Range settings on source and display |
| Blacks crushed and whites clipped | A full-range signal shown as narrow range | Range settings on source and display |
| Pink, green or purple picture | RGB and YCbCr mismatched | Colour format settings on source and sink |
| Picture cropped, stretched or in a box | Scaler or display aspect and overscan settings | Scaling and aspect settings along the chain |
| Motion judders or stutters | Frame-rate mismatch and conversion | Frame rates at every stage |
| Jagged combing on moving edges | Interlaced video shown without deinterlacing | Source format and the deinterlacer |
| Lips move after the sound | Video processing delay not matched in the audio | Total video delay, then delay the audio |
How to do it
- Write down what the signal should be at the source, the middle and the end.
- Send a known-good test pattern into the start of the path.
- Check the signal at the middle of the chain.
- Repeat in whichever half is wrong until one device or cable is left.
- Read the format each device reports at that point, and compare it with what it should be.
- Change one thing, check, and write it down.
Practice it
A laptop feeds a scaler, which feeds an HDBaseT link to a projector. The projector shows a grey, flat picture with no true black.
- You know what the signal should be at every stage before you start.
- You start from a known-good source.
- You split the path in halves rather than checking in order.
- You can name the likely cause from what the fault looks like.
- You change one thing at a time and write it down.
Toward Video Engineering
Working video knowledge and the path to specialist roles.
The intermediate video professional
Pulling it together, Video Intermediate develops working knowledge of video systems — signals (resolution, frame rate, colour), formats and standards (SDI, HDMI, IP, codecs), processing (conversion, scaling, sync), signal paths, and troubleshooting. The intermediate video professional understands video as signals moving through systems and can set up, connect, and troubleshoot professional video. This is the level that makes you genuinely capable with video systems and is the foundation for video engineering and specialized video roles (switching, engineering, playback, broadcast).
This builds on Video Basics toward the technical video roles across the industry (video engineer, switcher, broadcast, live video, playback), all of which rest on this systems knowledge. A capable intermediate video professional understands and works with video systems and is ready to specialize and advance. Understanding intermediate video is a key step in a technical video career.
Developing your video knowledge
Developing your video knowledge means working hands-on with real video systems and signals (setting up, connecting, converting, troubleshooting), deepening your understanding of signals, formats, and processing, and learning the gear (converters, scalers, switchers, etc.). From here you can specialize — video switching, video engineering, playback/media servers, broadcast video, AV-over-IP — and advance into technical video roles. The best video professionals have deep systems understanding, broad format and gear knowledge, and sharp troubleshooting, built through hands-on work. Deliberately developing these is the path forward.
From here, growing means working with more and more complex video systems, specializing in a video discipline, and building toward video engineering and senior technical video roles. Video is a deep, technical, in-demand field (especially as it moves to IP). The skills this book brings together — signals, formats, processing, paths, troubleshooting — are the foundation; mastery comes from hands-on systems work and specialization. Keep working with video systems, keep deepening your knowledge, and you grow toward video engineering.
What separates a senior video technician
Senior video technicians are the ones whose systems work the first time. They write down the format at every stage before they build, they choose cables for the data rate rather than for what is in the case, they manage EDID instead of hoping, they know where the HDCP will be before the client's laptop arrives, and they measure before they trust a monitor.
From here the paths divide. Video Engineering adds timing and camera matching and the running of a whole video system. Advanced Switching and Signal Routing adds the large switchers and routers. Projection and Display Systems goes deep on the screens. All of them assume the signal knowledge in this book.
Practice it
You are asked to plan the video for a conference with a 4K60 LED wall 40 metres from the control room, four presenter laptops and an NDI feed to a streaming team.
- You plan the format at every stage before building.
- You choose cable and transport by data rate and distance.
- You measure a signal before trusting a monitor.
Glossary
| Codec | How video is encoded/compressed for storage or transmission |
| Conversion | Changing a signal between formats |
| Frame rate | Frames (images) per second |
| Genlock | Synchronizing signals' timing |
| Resolution | The image's pixel detail (e.g., HD, 4K) |
| Scaling | Resizing the image to a different resolution |
| SDI | The professional video interface (over BNC/coax) |
| AV-over-IP | Carrying video over networks |
| 3G-SDI | SDI at 2.97 Gb/s, for 1080p at 50 and 60 frames; Level A maps the picture directly, Level B carries two HD links |
| 12G-SDI | SDI at 11.88 Gb/s, for UHD on one cable |
| Quad link | UHD split across four SDI cables, as square division or two-sample interleave |
| TMDS | The coding of HDMI up to version 2.0, three data pairs and a clock, ten bits sent for every eight |
| FRL | HDMI 2.1's fixed rate link, four data lanes with an embedded clock |
| EDID | The record a display keeps to tell sources what it can show |
| Hot plug detect | The line that tells a source a display has been connected |
| HDCP | Encryption between devices for protected content, with limits on devices and depth |
| Repeater | A device that passes HDCP on to further devices, such as a switcher or splitter |
| YCbCr | A brightness signal and two colour-difference signals |
| Narrow range | Black at 16 and white at 235 in 8-bit, 64 and 940 in 10-bit; the video default, also called limited range |
| Full range | Every code value used for picture, as computers usually send |
| Pixel clock | Samples per second including blanking; 148.5 MHz for 1080p60 |
| InfoFrame | A small HDMI packet describing the picture, including HDR metadata |
| VPID | The payload identifier in SDI that flags format, curve and colour space |
| Frame synchroniser | A device that stores frames to lock a source to system timing, adding delay |
| Deinterlacing | Turning two interlaced fields into progressive frames |
| HDBaseT | Uncompressed video, audio, USB, Ethernet, control and power over category cable |
| NDI | A compressed video-over-network format for ordinary networks |
| SRT | A protocol for sending compressed video reliably over the internet |
| SMPTE ST 2110 | Uncompressed video, audio and data as separate network streams locked to a common clock |
| PLUGE | Patches just below, at and above black, for setting a monitor's black |
| Vectorscope | A display of colour, with target boxes for colour bars |
| CRC error | A detected data error on an SDI link, the warning before it fails |
Where to Go Next
Video Engineering adds timing, camera matching and the running of a whole video system. Advanced Switching and Signal Routing adds the large switchers and routers. Projection and Display Systems goes deep on the screens. Camera Theory explains the picture's colour, range and sampling from the camera's side.
The first ninety days
Four things. Work out the pixel clock and bandwidth of every format you use, and check it against your cables. Set fixed EDID on every computer input of the next show you build. Learn to read colour bars on a waveform and vectorscope until a wrong one jumps out. And keep a list of every fault you find, what it looked like, and where it was.
What the gear actually looks like
Every device named in this book, in one place. Recognition is the whole point of Chapter 4 — use this page to attach a picture to each name before you meet it on a dock at six in the morning.
Find the destination.
Work along the path between them.