SMPTE Color Bars: The 70-Year-Old Test Pattern Still Running Your Signal Chain

There’s something to be said for a tool that outlives the technology it was built for. SMPTE color bars were designed to align analog NTSC monitors in an era of vacuum tubes and composite video. NTSC is dead. The bars are still here, still the first thing that goes down a new cable, and still the fastest way to find out whether your chain is lying to you.

I started paying real attention to them when I began mixing cameras, converters, and a switcher in the same rig. When a source looks slightly off and you can’t tell whether it’s the camera, the converter, the switcher, or the monitor, guessing costs you an hour. Bars cost you thirty seconds.

Why Color Bars Still Matter

The bars aren’t a calibration target in the modern colorimetry sense. They’re a known signal. That’s the whole point.

Every value in the pattern is defined. If you send bars into a chain and what comes out the other end doesn’t match what went in, something in the middle is doing something to your video — and the pattern is designed so that the kind of error tells you where it lives. Levels stretched? A range mismatch. Colors rotated on the vectorscope? An encoding problem. Black crushed? Someone’s monitor is set wrong, or a converter is doing a legal-to-full conversion nobody asked for.

In a live rig with a camera, a converter, a switcher, and a stream encoder, that’s four places for a setting to be wrong. Bars find it.

A Short History of SMPTE and the Bars

SMPTE itself predates television entirely. It was founded in 1916 as the Society of Motion Picture Engineers, back when the standardization problem was film gauges and projection speeds. Television got added to the name in 1950, and the organization has been writing the industry’s plumbing standards ever since — timecode, SDI, DCP packaging, and yes, test patterns.

The color bar pattern’s ancestor came from RCA Laboratories. Norbert D. Larky and David D. Holmes conceived it and first published it in RCA Licensee Bulletin LB-819 on February 7, 1951, with U.S. patent 2,742,525 for a Color Test Pattern Generator awarded to them on April 17, 1956. The EIA later published its own version, RS-189A, which became EIA-189A in 1976 as a standard color bar signal for adjusting monitors and encoders and running quick checks on transmission systems.

The pattern most people picture came out of CBS. In 1977, A. A. Goldberg of the CBS Technology Center described an improved color bar test signal developed there by Hank Mahler, and submitted it to the SMPTE TV Video Technology Committee. It was published as SMPTE ECR 1-1978. CBS never filed a patent on it, which put it in the public domain for the entire industry to use — a decision that’s a big part of why these bars became universal. The work picked up a Technology & Engineering Emmy in 2002.

SMPTE refined it into Engineering Guideline EG 1-1990, which is the document people mean when they say “SMPTE bars.” Then HD happened, and the pattern had to grow up.

The Classic: EG 1-1990

This is the one with the split field and the odd-looking strip in the middle.

It’s 4:3, built for NTSC, and organized in three horizontal bands:

Top two-thirds — seven bars at 75% amplitude: white, yellow, cyan, green, magenta, red, blue. Left to right in descending luminance order, which matters for how you read them on a waveform.

Middle strip — the reverse blue bars, sometimes called castellations: blue, black, magenta, black, cyan, black, white. This strip exists for one trick, which I’ll get to below.

Bottom band — the -I patch, 100% white, the +Q patch, and a black field containing PLUGE. The I and Q patches reference the NTSC color encoding axes; the white patch sets your white level; PLUGE sets your black.

Everything about this pattern assumes composite analog video. The I and Q patches only mean something in an NTSC encoding context. The 7.5 IRE setup pedestal it was built around doesn’t exist in digital component video.

The New: RP 219

SMPTE published RP 219 in 2002 for HD, revised it as RP 219-1 in 2014, and added RP 219-2 in 2016 covering 2048×1080, 3840×2160, 4096×2160, and 7680×4320.

It’s 16:9 and organized in four horizontal bands rather than three. The clever part is that it preserves a central 4:3 region containing the actual color bars, with 40% gray bars filling the 16:9 wings. Downconvert an RP 219 signal to SD and you get a familiar, correct SD pattern out the other end — which was the entire design goal in a world where everything was being converted between formats constantly.

Pattern 1 — 7/12 of frame height (630 px at 1080)

40% gray, the seven 75% bars, then 40% gray again. The gray wings are 240 px each and the seven color bars are 206 px each, which is what fills the central 4:3 region exactly.

Left to right: 75% white, yellow, cyan, green, magenta, red, blue — descending luminance order. On a waveform this reads as a clean staircase stepping down, and any bar out of sequence means something is badly wrong upstream. In full-range RGB the 75% bars measure 190; in 10-bit legal range they sit at 180.

Pattern 2 — 1/12 (90 px at 1080)

Four segments: 100% cyan (240 px), the +I patch (206 px, a dark navy that measures RGB 0/57/106), a long 75% white field (1236 px), and 100% blue on the right (238 px).

The 100% cyan and blue blocks sit directly beneath their 75% counterparts in pattern 1, which is the point — you can see 75% and 100% versions of the same hue stacked, and compare them at a glance.

Pattern 3 — 1/12 (90 px at 1080)

Mirror image of pattern 2 in structure: 100% yellow on the left (240 px), the +Q patch (206 px, a deep violet at RGB 66/11/122), the luminance ramp across the center, and 100% red on the right.

The ramp runs black to white across roughly 1236 px, which at 8-bit gives you about five pixels per code value — it’s a genuinely fine gradient. This is the band worth staring at. Flat patches can hide a bad conversion completely, but a ramp shows banding, posterization, and gamma kinks immediately. If you see stair-stepping instead of a smooth wash, you’re losing bit depth somewhere in the chain.

Pattern 4 — 3/12 (270 px at 1080)

Seven segments, left to right: 15% gray (240 px), 0% black, 100% white (412 px), 0% black, then the PLUGE steps, then 0% black and 15% gray to close.

Note the side bars drop to 15% gray here, not the 40% used in pattern 1. RP 219 sets 15% as the default value for these two blocks.

The PLUGE itself is the small stuff on the right, and it’s easy to miss because it’s supposed to be nearly invisible:

At normal levels those steps measure RGB 4 and RGB 10 against black — roughly +2% and +4%. That’s the whole trick. Your monitor is set correctly when the +2% bar is just barely distinguishable from the black around it.

One important caveat about still files. The spec calls for a -2% sub-black step, but sub-black doesn’t exist in a full-range RGB container — it clips to zero. If you’re using a PNG or TIFF of this pattern, that step is gone, and you can’t use it to check for black crush below zero. For that you need a real video-range signal: a generator, an SDI feed, or a 10-bit YUV file. Worth knowing before you conclude your monitor passed a test it never actually took.

RP 219-2 dropped the ±I and +Q options entirely for the UHD formats, on the reasoning that nobody’s converting a 4K signal to NTSC. You’ll still see them in 1920×1080 RP 219-1 output, which is why they’re in the examples above.

Classic vs New: What Actually Changed

• Aspect ratio — 4:3 becomes 16:9 with a preserved 4:3 core for downconversion
• The ramp — new to RP 219, and the best addition. Reveals banding and gamma errors that solid patches can’t
• PLUGE steps — classic uses -4/0/+4 IRE around a 7.5 IRE pedestal; RP 219 uses -2%, 0, +2%, +4% in digital percentages, plus sub-black
• I and Q — present in the classic for NTSC axis reference, gone from the UHD spec
• Reverse blue castellations — the classic’s middle strip is purpose-built for blue-only monitor alignment; RP 219 restructures this into the 100% color flanking blocks
• Bit depth expectations — RP 219 assumes 10-bit or better, since the ramp and near-black steps band visibly at 8-bit

If you’re working in HD or above, use RP 219. The classic pattern is for matching legacy gear, feeding SD equipment, or because you want the look.

How to Actually Use Them

Set black with PLUGE. Find the near-black bars at the bottom. On a correctly adjusted monitor, the above-black bar is just barely visible and the below-black bar is invisible — indistinguishable from the black around it. If you can see both, brightness is too high and you’re lifting your blacks. If you can see neither, brightness is too low and you’re crushing shadow detail. Adjust until that one bar just barely separates.

Set white with the 100% white patch. Raise contrast until the patch stops getting brighter or starts blooming, then back off. On an OLED or a modern reference monitor this matters less than it did on CRTs, but it’s still the right check.

Use blue-only mode for chroma and hue. This is the trick the classic pattern’s castellation strip was built for. Switch your monitor to blue-only. The top bars become alternating light and dark columns, and with the strip below them, four of the columns should read as continuous top to bottom — no visible seam where the bands meet. If a column has a brightness step in it, your chroma or hue is off. Adjust until the seams disappear. It works because blue is either fully present or fully absent in those specific patches, so any error shows up as a luminance mismatch your eye catches instantly.

Read the waveform. The 75% bars should sit at 180 in 10-bit legal range terms, not 255. If they’re at 255, something converted your levels. Black should sit at 64, not 0. This single check catches the most common problem in modern workflows.

Read the vectorscope. The six color bars should land in their targets. Rotation means a hue error; distance from center means saturation is off. On a chain with no problems, they sit in the boxes.

Check the ramp for banding. Any stair-stepping in what should be a smooth gradient means you’re losing bit depth somewhere — an 8-bit link, a bad conversion, or aggressive compression.

Bars and Tone

Bars almost never travel alone. The convention is bars plus a 1 kHz reference tone, laid at the head of a tape, a file, or a feed.

The tone level depends on where you are. North American broadcast practice generally puts the reference at -20 dBFS. EBU practice uses -18 dBFS. Both correspond to the same “0 VU” idea from the analog world, and confusing them costs you 2 dB across the whole program — which is exactly the sort of thing bars and tone exist to prevent. Pick the one your deliverable spec calls for and be consistent.

For a live show, thirty seconds of bars and tone on the program feed before you go live tells your engineer, your stream platform, and your recording that everything downstream is receiving what you think it’s sending.

Using Them in Practice

In DaVinci Resolve — bars live in the Effects Library under Generators. If you bring in a bars image file instead, watch your levels: PNG and TIFF are full-range RGB, so 75% white lands at 190 rather than the 180 you’d expect in video-legal range. Right-click the clip, set Data Levels to Full, and your scopes will read correctly. Get this wrong and you’ll chase a nonexistent problem for twenty minutes.

In a live chain — ffmpeg generates both patterns as a live source. ffmpeg -f lavfi -i smptehdbars=size=1920x1080:rate=59.94 gives you RP 219 bars, and smptebars gives you the classic. Pipe that to a DeckLink output, an SRT endpoint, or an RTMP push and you can validate an entire path before a camera is even powered on.

On a shoot — record ten seconds of bars at the head of a card when you’re matching multiple cameras. It costs nothing and it gives your colorist a reference that isn’t a guess.

What I Love

• A known signal beats a subjective impression every single time
• PLUGE is the fastest black-level check that exists, and it takes five seconds
• Blue-only alignment works on any monitor with the mode, no probe required
• The RP 219 ramp catches bit-depth and banding problems nothing else surfaces
• Free — the pattern is public domain and every NLE, switcher, and converter generates it
• Diagnoses where a fault lives, not just that one exists
• The 4:3-safe center region in RP 219 means one pattern survives downconversion
• Works identically whether you’re checking a $300 monitor or a $30,000 chain

Worth Knowing

Bars won’t calibrate a display to a color standard — that needs a probe and proper software. They tell you whether a signal is intact and whether a monitor is grossly misadjusted, which is a different and narrower job. They also say nothing useful about HDR; RP 219 is a standard dynamic range pattern, and HDR needs its own test signals entirely. And the blue-only trick depends on your monitor actually having a real blue-only mode rather than a simulated one.

None of that is a knock. It’s a signal integrity tool, not a colorimeter, and it’s the best one there is at that job.

Final Verdict: Are Bars Still Worth Your Time?

Yes, and the moment you’re running more than one device in a chain the answer stops being close.

Modern gear fails quietly. A converter silently reinterprets your levels, a switcher outputs the wrong range, a monitor ships with contrast at 90, and none of it announces itself. You just end up with footage that’s slightly wrong and no idea why. Thirty seconds of bars at the top of a session turns that into a solved problem.

A test pattern from 1978 with a lineage back to 1951 is still the first thing I put down a new cable. That should tell you something.

Pattern Reference: RP 219 at 1920×1080

• Side gray bars: 240 px wide, 40% gray
• Color bars: seven bars, 206 px each, filling the central 4:3 region
• Pattern 1 height: 630 px (7/12)
• Pattern 2 height: 90 px (1/12)
• Pattern 3 height: 90 px (1/12)
• Pattern 4 height: 270 px (3/12)
• Bar colors: 75% white, yellow, cyan, green, magenta, red, blue
• PLUGE steps: -2%, 0, +2%, 0, +4%, plus sub-black valley
• Legal range values: black at 64, 75% white at 180, 100% white at 940 (10-bit)
• Covered formats: 1920×1080 (RP 219-1); 2048×1080, 3840×2160, 4096×2160, 7680×4320 (RP 219-2)