How OLED and LCD Displays Are Made: From Glass Substrate to Finished Panel

Every LCD and OLED panel starts as a sheet of ultra-flat, alkali-free glass, usually made by drawing molten glass down both sides of a refractory trough so the two outer surfaces never touch anything. A thin-film transistor backplane is then patterned onto that glass, and the light-control layer goes on top: liquid crystal sealed against a colour filter for an LCD, or organic emitters evaporated through metal masks and sealed against moisture for an OLED. Polarisers, touch sensors, cover glass and a display driver chip are bonded on last, and that driver chip, the timing controller and, in a monitor, the scaler all run firmware that reports to your computer through EDID and DDC/CI.

This post walks the whole path one level deeper than our smartphone build and laptop build explainers, and it borrows the transistor fabrication background from how semiconductors are made. We cover the glass, the three main transistor families, LCD cell assembly, OLED deposition and encapsulation, LTPO, tandem OLED and QD-OLED, bonding, the firmware inside a display and what EDID and DDC/CI reveal to the host. Where a figure comes from a manufacturer, we say which document it came from and when. Panel makers treat most of their yields and recipes as trade secrets, so where no primary source publishes a number, we say so rather than guess.

Key takeaways

• Display glass is not window glass. It is alkali-free and, for most large panels, formed by the overflow fusion downdraw process, first patented in the 1960s, which leaves both outer surfaces untouched by tooling.
• "Gen 10.5" is a glass sheet size, not a display technology. Corning gave Gen 10.5 as 2,940 by 3,370 mm in December 2015; Sharp gave its Gen 10 sheet as 2,880 by 3,130 mm in October 2009. A bigger mother glass changes how many TV panels you can cut from one sheet.
• The backplane transistor family decides a lot: amorphous silicon is cheap and slow, LTPS is fast, and oxide (IGZO) leaks so little that a pixel can hold its image with very little refreshing. LTPO mixes LTPS and oxide transistors to allow very low refresh rates.
• An LCD is a light valve: a backlight, two polarisers, a liquid crystal layer a few micrometres thick and a colour filter. An OLED pixel makes its own light, which is why its organic layers must be sealed from water and oxygen almost perfectly.
• Tandem OLED stacks two emitting units per pixel; QD-OLED uses blue OLED light and converts part of it to red and green with quantum dots.
• A finished display carries firmware in the driver IC, the timing controller and, in monitors, the scaler. Monitors describe themselves to the host through EDID, which can include a serial number and a manufacturing week, and accept control commands over DDC/CI.
• Monitor firmware updates arrive through vendor tools, through Windows driver packages and, on Linux, through fwupd and the Linux Vendor Firmware Service for supported devices.

The whole journey in one table

The step table below is the map for the rest of the post. Yields at each stage are closely guarded by panel makers and are not published in any primary source we could find, so the table names the defect check at each stage instead of inventing a percentage.

What is the glass substrate, and why is display glass special?

The substrate is the sheet everything else is built on. In an LCD there are two of them, one for the transistors and one for the colour filter. In a rigid OLED there is one glass substrate under the transistors plus a glass or thin-film seal on top. Flexible OLEDs swap the glass for a polyimide film, but even then the polyimide is usually coated onto a glass carrier for processing and lifted off at the end.

Display glass has to do things window glass cannot. Window glass contains sodium. Sodium ions migrate under heat and electric fields, and in a transistor they would shift threshold voltages and wreck the circuit. So display glass is alkali-free, built on a boroaluminosilicate recipe. It also has to stay dimensionally stable through repeated heating during transistor fabrication, which is why glass makers sell grades marketed for "low thermal compaction". If a 3 metre sheet shrinks by even a few micrometres between two lithography steps, the second layer lands in the wrong place relative to the first.

The third requirement is flatness and surface quality. A liquid crystal layer is only a few micrometres thick, and photolithography on a sheet the size of a garage door needs the surface to sit within a very shallow focus window. A scratch or a ripple shows up as a visible defect in the finished panel.

How fusion-drawn display glass is made

The process most associated with display glass is the overflow fusion downdraw, often just called the fusion process. Its origin is a patent titled "Sheet forming apparatus" filed by Stuart M. Dockerty, US 3,338,696, with a priority date of 6 May 1964 and publication on 29 August 1967, according to the Google Patents record.

A later Corning patent, US 8,210,001, describes the process in plain terms. Molten glass is delivered into a trough at the top of a forming body called an isopipe, which has an upper trough-shaped part sitting on a lower wedge-shaped part. The glass overflows the two side walls, which the patent calls weirs, as two separate streams. The streams run down the outside of the wedge and meet at its bottom edge, the root, where they fuse into one ribbon. The patent calls this "a leading process developed and used by Corning" and notes that the ribbon "has pristine surfaces."

That last point is the whole reason the method matters. The faces of glass that touched the isopipe end up fused together in the middle of the sheet. The two outer faces have only ever touched air. Compare that with the float process used for window glass, where one face sits on a bath of molten tin. Float glass can be polished to display quality, and some display glass makers use float lines, but polishing adds a step and cost. Fusion glass comes off the draw ready for lithography without polishing.

Below the root, the ribbon is pulled downward by rollers that grip only its edges. The pulling speed and the glass flow set the thickness. Edge rollers are cooled to fight the ribbon's tendency to narrow as it is drawn. The ribbon then passes through a long controlled cooling zone. The Corning patent is in fact about this zone: it describes managing temperature so that thermal artefacts in the ribbon stay within limits that keep stress birefringence, called retardation, below a chosen quality level. Retardation matters because a display sits between polarisers, and stress in the glass rotates polarised light, which can show as uneven brightness.

At the bottom of the draw the ribbon is scored and snapped into sheets, the edges are ground to remove tiny cracks, and the sheets are washed, inspected and packed in racks for shipment to the panel plant. Because Gen 10.5 mother glass is close to ten square metres and well under a millimetre thick, glass makers often build next to the panel fab, as the Corning release below describes.

What does "Gen 8.5" or "Gen 10.5" mean?

A generation number is just a mother glass size. It says nothing about the transistor type, the liquid crystal mode or whether the line makes LCD or OLED. Two dates anchor the large end of the scale.

The Sharp release also gives the plant's capacity as 72,000 substrates per month, starting at 36,000. The Corning release says the Hefei facility was planned to support a panel maker's mass production of large TV panels "by the third quarter of 2018" and put the total investment at approximately 1.3 billion US dollars.

Worked example: why half a generation matters

You can see the economics with simple geometry. A 16:9 panel with a 65-inch diagonal has an active area of about 1,439 by 809 mm. A 75-inch panel is about 1,660 by 934 mm. These are active areas only; real cut plans add bezels, seal lines and gaps between panels, so treat the counts below as upper bounds.

• Gen 10 (2,880 by 3,130 mm), 65-inch. Lay panels with the long side along the 2,880 mm edge: two fit (2,878 mm), and along 3,130 mm three fit (2,427 mm). That is 6 panels. Rotating them gives 2 by 3 again, so 6.
• Gen 10.5 (2,940 by 3,370 mm), 65-inch. Put the long side along the 2,940 mm edge: two fit (2,878 mm). Along 3,370 mm, four short sides fit (4 times 809 is 3,236 mm). That is 8 panels.
• Gen 10.5, 75-inch. Two long sides need 3,320 mm, which fits along 3,370 mm, and three short sides need 2,802 mm, which fits along 2,940 mm. That is 6 panels.

Going from Gen 10 to Gen 10.5 adds about 10 percent more glass area (9.01 to 9.91 square metres) but, for 65-inch panels in this simple layout, a third more panels. The extra 60 mm on one edge and 240 mm on the other is exactly what lets a fourth row of 65-inch panels fit. That is why the Corning release ties Gen 10.5 to 65-inch and 75-inch TVs specifically. Smartphone and laptop OLED lines run on much smaller sheets, often called Gen 6, because evaporation masks get harder to hold flat and aligned as the substrate grows.

How is the thin-film transistor backplane built?

The backplane is a grid of transistors, one or more per sub-pixel, laid down on the glass. Each transistor acts as a switch, or for OLED as a current regulator, for one dot of colour. A 4K panel has about 8.3 million pixels and roughly 25 million sub-pixels, so the backplane is a chip-scale circuit stretched across a TV-sized sheet.

The method is the same family of steps used for chips, run at much lower resolution and much larger size:

1. Clean the glass.
2. Deposit a layer: metal for gate lines by sputtering, silicon nitride or silicon oxide insulators and the semiconductor by plasma-enhanced chemical vapour deposition, or an oxide semiconductor by sputtering.
3. Coat photoresist, expose it through a photomask using a large-area stepper or scanner, and develop it.
4. Etch away the exposed layer with wet chemicals or plasma, then strip the resist.
5. Repeat for the next layer. A simple amorphous silicon backplane may need four or five mask layers; LTPS backplanes for phones need many more.
6. Test the finished array electrically, then repair shorts or open lines with lasers where possible.

The semiconductor you choose for step 2 is the decision that sets everything else.

Amorphous silicon (a-Si)

Amorphous silicon was the workhorse of LCD TVs and monitors for decades. It deposits at low temperature over very large areas and needs few mask steps, so it is cheap. Its weakness is low electron mobility, which means the transistors must be relatively large to pass enough current. That limits pixel density and makes a-Si a poor fit for driving OLED pixels, which need a steady current rather than just a voltage switch.

Low-temperature polysilicon (LTPS)

LTPS starts as amorphous silicon and is crystallised with a pulsed excimer laser that briefly melts the film without heating the glass enough to deform it. The resulting polycrystalline silicon has far higher mobility, so transistors are small and fast. That enables high pixel density and lets designers put some drive circuitry, such as gate drivers, directly on the glass. LTPS is the standard backplane for phone OLEDs. Its drawbacks are cost, many mask steps, uniformity limits from the laser scan on very large sheets, and relatively high leakage current when the transistor is switched off.

Oxide TFTs (IGZO)

Oxide TFTs use a metal oxide semiconductor, most commonly indium gallium zinc oxide (IGZO), deposited by sputtering. Sharp's own IGZO page states that IGZO's electron mobility is "20-50 times greater" than the amorphous silicon used in many LCD panels. It also explains the feature that matters most for power: an IGZO pixel can hold a still image without being rewritten, which Sharp says reduces power consumption to "a fifth or even a tenth" of conventional screens, and notes that the first smartphone with its new-generation IGZO display launched in Japan in November 2012. Oxide also scales to large sheets better than laser-crystallised silicon, which is why it is used in large OLED TV backplanes.

Why the choice matters for refresh rate and power

The connection between leakage and refresh rate is direct. A pixel stores its brightness as charge on a small capacitor. Between refreshes, that charge drains through the switching transistor's off-state leakage. If leakage is high, the display has to refresh often or the image drifts and flickers. If leakage is very low, the display can refresh rarely, and every refresh you skip saves the power spent driving the data lines. That is the whole idea behind LTPO, covered below.

How are LCD cells assembled?

An LCD makes no light of its own. It is a valve that lets more or less backlight through each sub-pixel. The standard four-stage flow is: TFT array, colour filter, cell assembly, and module.

The colour filter

The colour filter is built on the second glass sheet. First comes a black matrix, a grid of opaque lines that hides the transistors and wiring and stops colour bleeding between sub-pixels. Then red, green and blue pigmented resists are coated and patterned one at a time by photolithography. A transparent overcoat smooths the surface, and many designs pattern small photo spacers, tiny pillars that hold the two glass sheets a precise distance apart.

Alignment, sealing and filling

Both inner surfaces receive a thin polyimide alignment layer. Liquid crystal molecules need a preferred direction at each surface, and the alignment layer provides it. Traditionally this was done by rubbing the layer with a cloth roller; newer lines use photo-alignment with polarised ultraviolet light, which avoids dust and scratches. Sharp's 2009 Sakai release names its own photo-alignment technology as part of the plant's design.

The modern way to fill the cell is one drop fill. A sealant line is drawn around each panel's edge on one substrate. Precisely measured droplets of liquid crystal are dispensed inside the seal. The two substrates are aligned and pressed together in a vacuum chamber so no air is trapped, and the seal is cured with ultraviolet light and heat. Older lines filled cells slowly by capillary action through a gap in the seal; one drop fill is much faster and wastes less liquid crystal.

The gap between the plates, set by the spacers, is only a few micrometres. Variation in that gap changes how much each region rotates the light, which shows up as mura, the uneven cloudy patches you sometimes see on a grey screen. Cell gap inspection is one of the main quality gates.

Polarisers

A polariser goes on each outer surface. The rear polariser makes the backlight polarised; the liquid crystal rotates or does not rotate that polarisation depending on the voltage at each pixel; the front polariser then passes or blocks it. Polarisers are laminated films, typically a stretched iodine-doped polyvinyl alcohol layer between protective films, and each one absorbs more than half of unpolarised light. That is one reason LCDs need bright backlights.

The backlight

Behind the cell sits the backlight unit, and there are two main designs.

• Edge-lit. A row of LEDs along one or two edges shines into a thin acrylic or glass light guide plate printed or etched with a dot pattern that scatters light forward evenly. Diffuser and prism films on top spread the light and push it toward the viewer. Edge-lit backlights are thin and cheap but can only dim in coarse strips, if at all.
• Direct-lit and mini-LED. LEDs sit in a grid directly behind the panel. With mini-LED the LEDs are small and numerous, often thousands, grouped into hundreds or thousands of separately dimmable zones. The timing controller or a dedicated local dimming processor analyses each frame and dims zones behind dark content. This raises contrast sharply, though small bright objects on black can still show a halo because one zone covers many pixels.

Many LCD backlights use blue LEDs plus a quantum dot film or phosphor to make a white with purer red and green, which widens the colour gamut. That idea, converting blue to red and green with quantum dots, comes back in QD-OLED below.

How are OLED emitters deposited?

The US Department of Energy describes an OLED as a solid-state device with a thin, carbon-based semiconductor layer that emits light when current is applied through adjacent electrodes. At least one electrode must be transparent so the light can escape. The brightness depends on the current and the colour depends on the emissive material. The DOE page is written about OLED lighting panels, but the physics is the same for display pixels.

A display OLED stack is a sandwich of very thin layers between an anode and a cathode: a hole injection layer, a hole transport layer, the emissive layer, an electron transport layer and an electron injection layer. Each is tens of nanometres thick. The whole stack is far thinner than a human hair. The emissive layer is what changes from red to green to blue sub-pixels.

Fine metal mask evaporation

For phones, watches and most small OLEDs, the organic layers are deposited by vacuum thermal evaporation. Organic material is heated in a crucible inside a high-vacuum chamber until it sublimes. The vapour travels upward and condenses on the substrate, which hangs face down above the sources.

To put red material only on red sub-pixels, a fine metal mask (FMM) is held against the substrate. The mask is a thin sheet with an opening for every sub-pixel of one colour. A Google Patents record for a fine metal mask patent, US 9,188,856, says the masks are commonly produced from Invar, a nickel iron alloy chosen because it barely expands with heat, by coating the sheet with photoresist, exposing the pattern, developing it and chemically etching the openings. The mask sheets are stretched and welded onto a rigid frame so they stay flat.

The hard parts are alignment and sag. The mask must line up with the backplane to within a few micrometres across the whole sheet, and a thin metal sheet held horizontally sags under its own weight. Each colour needs its own mask and its own pass. Masks also collect organic material and need regular cleaning. These limits are a big part of why small OLED lines use smaller mother glass, and why higher pixel densities push mask makers toward thinner sheets and other manufacturing methods such as electroforming.

The common layers, such as the transport layers and the cathode, go down through an open mask that covers the whole active area. The cathode is often a thin semi-transparent metal for top-emitting designs, where light exits through the top of the stack rather than through the backplane.

White OLED with colour filters

Large OLED TVs have historically sidestepped the fine metal mask problem with a different architecture: every sub-pixel emits white light from a stack of emitters deposited without a fine mask, and a colour filter on top picks out red, green or blue, often with an unfiltered white sub-pixel added for brightness. This loses some light in the filters but makes very large panels practical.

Inkjet printing

The alternative to evaporation is to print. In inkjet printing, the organic material or another functional material is dissolved in a solvent and jetted as tiny droplets into wells on the backplane, called banks, that keep each droplet in its own sub-pixel. As the solvent dries, the film forms. Printing uses material only where it is needed, avoids masks, and scales to large sheets. The challenges are making films with uniform thickness as they dry, finding solvents that do not attack the layer underneath, and matching the lifetime of evaporated materials, especially for blue. Printing is used today for some layers, notably the quantum dot colour conversion layer in QD-OLED, according to industry reporting, while evaporation still dominates the emissive layers in mainstream phone OLED.

Why encapsulation is so critical

Organic emitters and the reactive metals used in OLED cathodes degrade quickly when they meet water vapour or oxygen. The visible symptom is dark spots that grow over time from pinholes, and dimming at the panel edges. An LCD can tolerate a slightly imperfect seal; an OLED cannot. The barrier has to be many orders of magnitude tighter than ordinary food packaging film, which is why encapsulation is treated as one of the hardest steps in OLED manufacturing.

There are two main approaches:

• Glass lid with frit seal. A second glass sheet is placed over the panel and sealed at the edges with a glass frit paste that is melted with a laser. This is a near-perfect barrier but rigid, so it suits flat rigid panels.
• Thin-film encapsulation. Alternating layers of inorganic barrier material, such as silicon nitride or aluminium oxide, and a planarising organic layer are deposited directly on the cathode. The inorganic layers block water; the organic layer covers particles and pinholes so a defect in one inorganic layer does not line up with a defect in the next. Thin-film encapsulation is what makes flexible and foldable OLEDs possible, because a glass lid cannot bend.

Panels are then qualified with accelerated tests at high temperature and humidity, and manufacturers watch for dark spot growth. IEC 62341-1-1:2009, the generic specification for OLED displays from IEC Technical Committee 110, sets general rules for electrical and optical measurements, environmental and mechanical tests and endurance tests within the IECQ quality assessment system. The IEC webstore lists a publication date of 7 April 2009 and a stability date of 2029.

What are LTPO, tandem OLED and QD-OLED?

These three names describe different fixes for different problems. LTPO is a backplane change aimed at power. Tandem OLED is an emitter stack change aimed at brightness and lifetime. QD-OLED is a colour architecture change aimed at colour volume and large-panel manufacturing.

LTPO: low-temperature polycrystalline oxide

LTPO combines LTPS transistors and oxide transistors on the same backplane. LTPS transistors handle the jobs that need speed and current; oxide transistors handle the switches that must hold a pixel's charge with very little leakage. The result is a panel that can refresh fast when content moves and very slowly when it does not.

Apple's newsroom release for Apple Watch Series 5, published in September 2019, describes its always-on display as using "the industry's only low-temperature polysilicon and oxide display (LTPO)," together with an ultra-low power display driver and a power management chip. Apple's September 2021 release for iPhone 13 Pro describes ProMotion as "an adaptive refresh rate from 10Hz to 120Hz," and the current iPad Pro technical specifications list ProMotion with "adaptive refresh rates from 10Hz to 120Hz."

Worked example: what refresh rate does to drive power

The power spent charging the data lines on a panel scales roughly with how many times per second they are charged. Take a simplified panel where data line driving costs a fixed amount of energy per frame, call it E.

• At 120 Hz, the panel spends 120 E per second.
• At 60 Hz, it spends 60 E, half as much.
• At 10 Hz, it spends 10 E, one twelfth of the 120 Hz figure.
• At 1 Hz, it spends 1 E, one hundred and twentieth.

Real panels do not scale that cleanly. The backlight or OLED emission, the timing controller, the touch controller and fixed driver overheads do not shrink with refresh rate, and very low rates need extra circuitry to stop flicker. But the arithmetic shows why a pixel that can hold its charge, which is what oxide transistors provide, matters so much for a watch showing a static face all day or a phone showing a static page of text.

Tandem OLED

A standard OLED has one emitting unit between the anode and cathode. A tandem OLED stacks two (or more) emitting units in series, connected by a charge generation layer. Each unit emits light, so for the same current the stack produces roughly twice the light, at roughly twice the voltage. Put the other way, the panel can reach the same brightness with about half the current density through each unit, and organic emitters age more slowly at lower current density. The trade is extra layers and higher drive voltage.

Apple's iPad Pro specifications page lists an "Ultra Retina XDR display" with "Tandem OLED," quoting 1000 nits maximum full screen brightness for SDR and XDR content and 1600 nits peak for HDR content only. Those are Apple's figures for that product; the general point is that tandem stacks let OLED reach high brightness without burning out the emitters.

QD-OLED

QD-OLED replaces the white OLED plus colour filter approach for large panels. Every sub-pixel emits blue from an OLED backplane. On top sits a colour conversion layer: red sub-pixels are covered with red quantum dots, green sub-pixels with green quantum dots, and blue sub-pixels pass the blue light straight through. Quantum dots are semiconductor nanocrystals that absorb a photon and re-emit at a wavelength set by their size, with a narrow spectrum.

An IEEE Spectrum teardown of a 55-inch QD-OLED TV published in 2022 describes exactly this split: blue pixels take their light directly from the OLED, while red and green are lit by quantum dots. The author found the green quantum dots arranged in their own line rather than in an even grid, observed that the native quantum dot emission was so narrow that the TV sometimes compresses its gamut for content mastered for narrower colour spaces, and noted that the quantum dot front plane separated from the OLED back plane surprisingly easily once the panel was opened. A companion Spectrum article explains the motivation: OLED emitters of different colours age at different rates, and using one blue emitter plus conversion aims to keep colours accurate as the panel ages.

How do touch sensors, cover glass and the driver chips get bonded on?

Once the panel is sealed and cut from the mother glass, it is still just a glass sandwich with tiny pads along one edge. The next steps connect it to the outside world.

Cutting and polarisers

The mother glass is scribed with a diamond wheel or a laser and broken into individual panels. For phones, laser cutting also creates notches, holes for front cameras and rounded corners. Polarisers are laminated next. OLED panels usually carry a circular polariser on the front, not to control the light but to absorb reflections from the metal layers inside, which would otherwise wash out blacks in daylight. Some newer OLED designs replace the polariser with a colour filter on encapsulation to recover the light that the polariser absorbs.

The display driver IC and timing controller

A display driver IC (DDIC) converts digital pixel values into the analogue voltages or currents applied to each column of the panel, and drives the scan lines row by row. A timing controller (TCON) receives the video stream from the host, reorders it, applies corrections and feeds the driver ICs. In phones the TCON functions are usually integrated into a single DDIC. In TVs and monitors there are several source driver chips along the top edge of the panel and a separate TCON on a small board.

Driver chips are attached in one of three ways:

• Chip on glass (COG). The bare driver IC is bonded directly onto a ledge of the glass substrate. This is compact but needs a strip of glass beyond the active area, which widens the bottom bezel.
• Chip on film (COF). The driver IC is bonded to a flexible film which is then bonded to the glass edge. The film can fold behind the panel, shrinking the bezel.
• Chip on plastic (COP). For flexible OLED on polyimide, the panel's own substrate is extended and bent back under the display, with the driver bonded on that extension. That is how phones achieve a thin bottom border.

The bond itself usually uses anisotropic conductive film (ACF): an adhesive loaded with tiny conductive particles. When the chip or film is pressed down with heat, particles trapped between each pad and its partner make an electrical connection vertically, while particles between neighbouring pads stay separated, so there is no short sideways. The panel is then lit up with test patterns to check every line.

Touch sensors and cover glass

Capacitive touch is a grid of transparent electrodes that senses the change in capacitance when a finger is near. There are three common placements:

• Out-cell or add-on. A separate touch sensor film or glass is laminated on top of the display.
• On-cell. The touch electrodes are patterned directly on top of the display, for OLED often on the thin-film encapsulation layer itself.
• In-cell. The touch function is built into the display's own electrode layers, common in LCDs, where the common electrode can be time-shared between display and touch.

The cover glass is a chemically strengthened aluminosilicate glass. Strengthening happens in a hot salt bath where larger potassium ions swap places with smaller sodium ions near the surface, packing the surface into compression so cracks struggle to start. Cover glass is laminated to the display with optically clear adhesive or a liquid adhesive cured by ultraviolet light. Bonding with no air gap reduces reflections between layers and improves readability in sunlight. Bubbles and dust trapped here are a costly late-stage defect.

Where does firmware live in a display?

This is the part most people do not think about. A modern display contains several programmable chips, and each one runs code or holds configuration data that can be updated, misconfigured or attacked.

Monitor firmware updates in practice

Monitor makers traditionally shipped firmware updates as Windows utilities that talk to the monitor over a USB upstream cable or over the video cable. On Windows, Microsoft also lets vendors deliver a corrected monitor description through an INF driver package, as covered in the EDID section below.

On Linux, the fwupd project and the Linux Vendor Firmware Service (LVFS) cover a growing set of display-related chips. The LVFS documentation explains that vendors upload firmware with metadata, and fwupd on the client queries supported hardware for its current firmware version and applies updates that vendors have published. Two fwupd plugins are directly relevant:

• The fwupd mediatek-scaler plugin documentation describes updating the firmware of DisplayPort display controllers. It says these devices communicate over I2C via the DisplayPort aux channel, are identified with a vendor-specific DDC/CI command, and that the firmware blob "is written to the partition of the device flash that is not currently running." Writing to the inactive bank is what makes a failed update recoverable. The same page notes that from fwupd 2.0.0 the device's instance ID is built from the manufacturer ID and product code found in the monitor's EDID.
• The fwupd synaptics-mst plugin supports querying and flashing multi-stream transport hubs used in systems and docks, the chips that split one DisplayPort signal into several monitors.

The practical lesson: a monitor has a firmware version, a vendor and an update channel, so it belongs in your hardware inventory like a dock.

What do EDID and DDC/CI expose to the host?

EDID: the display's identity card

Extended Display Identification Data is how a display tells the host what it is and what it can do. Microsoft's documentation describes EDID as a metadata format that monitors use to communicate their capabilities to a video source; it contains the monitor identifier, manufacturer data, hardware identifier and timing information, and is stored in the monitor's EEPROM in a VESA-specified format. Microsoft also notes that "All monitors, analog or digital, must support EDID." EDID data comes in 128-byte blocks: versions 1.0 to 1.2 are one block, while version 1.3, also called enhanced EDID or E-EDID, adds extension blocks.

VESA lists DDC/CI 1.1, DisplayID, E-DDC 1.2, the E-EDID releases and MCCS 2.2a among the standards it makes available free of charge. DisplayID is VESA's newer, more flexible structure for the same job, and modern displays may carry both.

The Linux kernel documentation shows what a decoded EDID looks like. Its EDID helper reference includes a full decode of a real 27-inch class monitor, produced by the edid-decode tool, listing the three-letter manufacturer code, the model number, a numeric serial number, "Made in: week 7 of 2022", the panel's physical size, its colour primaries, every supported timing and a separate "Display Product Serial Number" text string. The same reference describes how the kernel builds a 32-bit panel ID from the EDID's 16-bit manufacturer ID and 16-bit product ID to match panels that need special handling.

What that means for privacy

• Fingerprinting. A serial number plus a manufacturing week is stable, per-unit data. Any software on the host that can read EDID, which on most systems includes ordinary user-space programs, can read it. Combined with other hardware details, it can help identify a specific machine or desk setup. Treat EDID dumps in bug reports, forums and support tickets as containing a hardware serial number.
• Asset tracking. The same property is useful for IT teams. Monitor serial numbers from EDID let inventory tools record which display is attached to which machine without anyone reading a label.
• Leakage through docks and KVMs. The kernel documentation notes that a KVM switch may send its own EDID instead of querying the connected monitor. That can confuse inventory, and it means the EDID a host sees is not always the monitor's own.

What that means for integrity

EDID is not signed. The host trusts whatever it receives, which is why the operating systems provide override mechanisms when a monitor reports something wrong:

• The Linux kernel documentation explains that the CONFIG_DRM_LOAD_EDID_FIRMWARE option allows a corrected EDID file to be loaded from /lib/firmware, and the DRM helper reference says an EDID can be overridden through debugfs override_edid or the drm.edid_firmware parameter, in that priority order, and that "Having either of them bypasses actual EDID reads."
• Microsoft documents an INF-based override that lets vendors replace one or more 128-byte EDID blocks; the monitor driver then reads the corrected blocks from the registry and the rest from the EEPROM.

Malformed EDID has historically been a source of parser bugs in graphics drivers, because EDID parsing runs early, with high privilege, on data supplied by an external device. That is a reason to keep graphics drivers and kernels updated even on machines that never touch the internet, and to be cautious with untrusted adapters, capture devices and KVMs.

DDC/CI: the control channel

DDC/CI (Display Data Channel Command Interface) runs over the same I2C wires as EDID and lets the host change monitor settings. The ddcutil documentation describes it plainly: ddcutil is a Linux program for managing monitor settings such as brightness, colour levels and input source, and "Generally speaking, any setting that can be changed by pressing buttons on the monitor can be modified." It uses DDC/CI to talk to monitors that implement the Monitor Control Command Set (MCCS) over I2C, normally exposed as /dev/i2c-n devices, with support for some monitors that implement MCCS over USB instead. It does not support laptop panels, which are controlled through a different interface.

The ddcutil FAQ is also a useful reality check on how loosely the standard is followed. It notes that no MCCS value for a USB-C input exists, so manufacturers pick their own, and that capability strings reported by monitors can be wrong, giving an example where a monitor's advertised input codes did not match the codes it actually accepts.

From a security standpoint, DDC/CI means local software can change what a monitor shows (switch inputs, change brightness, toggle settings) and, as the fwupd documentation shows, some monitors accept vendor-specific commands over the same channel that are used for firmware updates. Access on Linux is governed by permissions on the /dev/i2c-* device nodes, which is why ddcutil's setup guide spends time on the i2c-dev driver and group permissions. Treat write access to those nodes like any other hardware access: grant it to the users who need it, not to everyone.

Step by step: read your own display's identity, settings and firmware

These steps are read-only unless marked. They work on a typical Linux desktop with a monitor connected over HDMI or DisplayPort, with Windows and macOS equivalents where they exist.

1. List connectors and find the one in use (Linux)

# Every connector the GPU driver exposes, with connection status
for c in /sys/class/drm/card*-*; do
  printf '%s: %s\n' "$(basename "$c")" "$(cat "$c/status" 2>/dev/null)"
done

2. Dump and decode the EDID

# Install the decoder (package name on Debian and Ubuntu)
sudo apt install edid-decode

# Decode the EDID for one connected output; replace the connector name
edid-decode /sys/class/drm/card1-DP-1/edid

Look for the manufacturer code, model, serial number, the "Made in" week and year, and the extension blocks. If you are about to paste this output in a public bug report, remove the serial number lines first.

3. Talk to the monitor over DDC/CI

sudo apt install ddcutil

# Load the I2C device driver if it is not loaded already
sudo modprobe i2c-dev

# Which monitors answer on DDC/CI, and on which I2C bus
ddcutil detect

# The monitor's self-reported feature list (MCCS capabilities string)
ddcutil capabilities --display 1

# Read current brightness (VCP feature code 10)
ddcutil getvcp 10 --display 1

ddcutil environment checks whether i2c-dev is loaded and whether your user can access the buses, which fixes most "no monitors found" problems. Many monitors also report a firmware level through an MCCS feature, which you can try with ddcutil getvcp c9 --display 1; not every monitor implements it, so an "unsupported feature" reply is normal.

Writing settings, for example ddcutil setvcp 10 50 --display 1 to set brightness to 50, is harmless on its own, but changing the input source (feature 60) can leave you with a blank screen until you switch it back with the monitor's buttons. Check the display number in ddcutil detect first if you have more than one monitor.

4. Check whether fwupd knows about your display chips

# Devices fwupd recognises, including monitor scalers, docks and MST hubs
fwupdmgr get-devices

# Refresh metadata from LVFS and see whether updates are published
fwupdmgr refresh
fwupdmgr get-updates

Warning: firmware flashing can brick a monitor or dock. Before running any update, confirm the device. Run fwupdmgr get-devices and note the exact device name and its device ID, check that the vendor and model match the monitor or dock on your desk, connect it directly rather than through another hub, and do not unplug power or cables during the update.

# Only after confirming the device ID from get-devices above
fwupdmgr update <device-id>

5. Windows equivalent

# Monitor identity from WMI (manufacturer, product code, serial, year of manufacture)
Get-CimInstance -Namespace root\wmi -ClassName WmiMonitorID |
  ForEach-Object {
    [pscustomobject]@{
      Manufacturer = -join ($_.ManufacturerName | Where-Object { $_ } | ForEach-Object { [char]$_ })
      Product      = -join ($_.ProductCodeID   | Where-Object { $_ } | ForEach-Object { [char]$_ })
      Serial       = -join ($_.SerialNumberID  | Where-Object { $_ } | ForEach-Object { [char]$_ })
      Year         = $_.YearOfManufacture
    }
  }

6. macOS equivalent

# Display model, resolution and connection type
system_profiler SPDisplaysDataType

# Raw EDID bytes for attached displays (output is long; search for EDID)
ioreg -lw0 | grep -i edid

7. Record what you found

Add each monitor's manufacturer code, model, serial and firmware version to your hardware inventory next to the computer it is attached to. That gives you a way to answer "which monitors are on firmware X" the day a vendor publishes a fix.

Common mistakes

• Treating "Gen 10.5" as a display technology. It is a glass sheet size. An LCD line and an OLED line can both be any generation.
• Assuming all OLEDs are made the same way. A phone OLED (LTPS backplane, fine metal mask evaporation, thin-film encapsulation) and a large TV OLED (oxide backplane, white OLED with colour filters or blue OLED with quantum dots) share a name and little else on the production line.
• Reading "mini-LED" as a self-emitting display. Mini-LED TVs and monitors are LCDs with a finely zoned backlight. Micro-LED, where each sub-pixel is its own LED, is a different and much rarer technology.
• Believing a higher refresh number always costs more power. With LTPO and oxide backplanes, a display that can drop its refresh rate on static content can use less power than a fixed 60 Hz display.
• Posting full EDID dumps in public tickets. They contain the monitor's serial number and manufacturing week.
• Assuming the EDID you see is the monitor's. KVMs, docks, splitters and capture devices may send their own.
• Ignoring monitor and dock firmware in inventory. Scalers, MST hubs and USB-C controllers are updatable chips with vendors and versions, and they ship fixes.
• Flashing firmware through a hub or over a flaky cable. Connect directly, confirm the device ID first and keep power stable.
• Granting everyone write access to I2C devices to make a brightness app work. Use a dedicated group for the users who need it.

FAQ

What is the difference between how OLED and LCD screens are made?

Both start with the same kind of glass and a thin-film transistor backplane. An LCD then adds a colour filter glass, a liquid crystal layer sealed between the two sheets, polarisers and a separate backlight. An OLED instead deposits organic light-emitting layers directly on the backplane and seals them against moisture, so it needs no backlight and no liquid crystal.

What does Gen 10.5 mean in display manufacturing?

It is the size of the mother glass sheet. Corning's December 2015 announcement gave Gen 10.5 as 2,940 by 3,370 mm, sized for economical cuts of 65-inch and 75-inch TV panels. The number says nothing about whether the line makes LCD or OLED.

Why is display glass made by the fusion process?

The fusion downdraw process forms the sheet in mid-air, so neither outer surface touches a roller or a tin bath while it is soft. The result is a flat, smooth surface ready for photolithography without polishing.

What is the difference between LTPS, IGZO and LTPO?

LTPS is laser-crystallised silicon, fast but leakier. IGZO is an oxide semiconductor with lower leakage, which lets pixels hold their image with less refreshing. LTPO puts both on the same backplane, using LTPS where speed matters and oxide where holding charge matters, so the display can vary its refresh rate down to very low rates.

Why do OLED screens need encapsulation?

The organic layers and the reactive cathode metal degrade when exposed to water vapour and oxygen, causing dark spots and dimming. Encapsulation seals them with a glass lid and frit or with stacked thin barrier films. Thin-film encapsulation is what allows flexible and foldable OLEDs.

What is the difference between QD-OLED and tandem OLED?

QD-OLED uses blue OLED emitters and converts part of the light to red and green with quantum dots. Tandem OLED stacks two emitting units in each pixel so it can reach high brightness at lower current per unit, which helps lifetime. They solve different problems and could in principle be combined.

Does my monitor have firmware?

Yes. The scaler on the main board runs the on-screen menu, input switching and DDC/CI, and the timing controller and driver chips have their own code or configuration. Many monitors also include USB hub and USB-C power delivery chips with firmware. Some of these can be updated with vendor tools, and on Linux some are supported by fwupd.

What information does EDID reveal about my monitor?

EDID includes a manufacturer code, model number, often a serial number and the week and year of manufacture, plus the screen size, colour characteristics and supported timings. Any software on your computer that can read display information can usually read it.

Is DDC/CI a security risk?

DDC/CI lets local software change monitor settings and, on some monitors, carry vendor commands used for firmware updates. It is a local channel, not a network one, so the risk is mainly about which local users and programs can access the I2C devices. Restrict that access and keep monitor firmware from trusted update channels.

How are touchscreens added to displays?

Either as a separate sensor laminated on top (out-cell), patterned directly on the display's top surface (on-cell, often on the OLED encapsulation), or built into the display's own electrodes (in-cell). A chemically strengthened cover glass is then bonded on with optically clear adhesive.

Where Safeguard fits, honestly

Safeguard does not make, inspect or test displays, and it does not scan monitor firmware, analyse hardware or generate a hardware bill of materials. Its models are single-purpose security models for software, so nothing in this post about glass, transistors or emitters is something Safeguard analyses. The fit is narrower and real: the software around displays is ordinary software. Monitor control tools, colour calibration apps, kiosk and digital signage images, fleet inventory agents that read EDID, and the Linux images that ship fwupd, ddcutil and graphics drivers all pull in open source dependencies. Safeguard scans repositories, container images and SBOM files for known vulnerable components, shows which projects carry a given component and version, and can open fix pull requests when a fixed version exists, with autonomous remediation switched off by default.

Rebuilt kiosks, signage players and desktop images pull the same graphics stack, monitor tools and update clients again every time they are built, and those open source components carry their own known vulnerabilities. The fit here is indirect, since Safeguard does not touch display hardware or firmware, but the software that manages displays is part of your supply chain like everything else.

Fix it without waiting for upstream

Whatever the vulnerability, most of the risk sits in two gaps. The first is the time between a fix being published and that fix actually running in your production builds. The second, and worse, is the time before any fix exists at all, while a maintainer works out a patch and a release schedule. Waiting through both gaps is a habit, not a requirement. Safeguard gives teams three ways to close them.

1. Autonomous remediation: pull requests, not tickets

When a fixed version exists, Safeguard's Griffin models open a tested, policy gated pull request that moves the affected dependency to the safe version. You can run it on one repository or in bulk across every project that carries the vulnerable version. Remediation covers eight categories, including dependencies, licenses, security posture and findings. Reachability analysis ranks the vulnerable code paths your application actually calls first, so the first pull requests you merge are the ones that matter most. Autonomous remediation ships switched off, with every category set to Manual, so your team decides how much runs on its own and what still needs a human approval.

2. Gold Open Source: the safe version in seconds, free

gold.safeguard.sh is free and needs no login. Search any CVE or package to see affected and fixed versions, CVSS, EPSS, CISA KEV status and a recommended safe version, or use the "Am I affected?" checker. The same data is available through a free read API, README status badges and RSS threat feeds. The safeguard-sh/gold-scan-action GitHub Action can fail a build on a known exploited vulnerability, on any known vulnerability, or never, so a version you have just removed cannot quietly come back in next week's merge.

3. Gold Registry: start from components that are already fixed

registry.safeguard.sh provides hardened, drop in replacements for upstream packages and container images. Every Gold artifact is held continuously to zero exploitable critical or high CVEs, zero malware and a completed license check, with SLSA L3 or higher provenance, a Sigstore signature and a public evidence page you can hand to an auditor. An artifact loses its Gold badge if it regresses. Teams can point their own package directory at the registry so hardened equivalents resolve without changing how developers install anything.

When upstream has no fix yet, or will not ship one, Safeguard first works to get the fix accepted upstream. If the issue is severe or upstream will not move, Safeguard builds a fork, fixes the vulnerability there and publishes a hardened Gold build, so you are not stuck on a maintainer's release schedule. You can also request a hardened, compatibility tested build of a specific package. This fork and patch flow is in early access today, so ask which packages it covers before you plan around it.

One honest limit: when the vulnerable code lives in a vendor appliance, a hosted service or an operating system runtime you do not build yourself, the fix still has to come from that vendor or distribution. The steps above apply to the open source dependencies and container images you build and ship, which is where most teams carry the largest and least tracked share of their exposure.

See how Safeguard would help

The first demo is a real Safeguard CLI session: it installs the CLI, opens the interactive terminal, and asks it to fix a lodash CVE in a repository. It reads the finding and advisory, asks before each command, edits package.json, runs npm install and npm test with approval and pushes nothing. The same flow applies to a monitor control tool, a calibration app or a signage player you maintain: scan the repository, then let the assistant propose the dependency upgrade under your approval.

Click through it at your own pace, or open it full screen.

The second demo sets up the Safeguard runner on your own machine, approves it with its device code and then scans from the terminal. That suits teams who build kiosk or desktop images on their own hardware and want the scan to run where the build runs, without handing over repository access.

Click through it at your own pace, or open it full screen.

The third demo connects Claude to the Safeguard MCP server with OAuth and chooses which tools Claude can use. If your team already asks Claude questions about its fleet, this is how the same conversation can look up which projects carry a vulnerable component.

Click through it at your own pace, or open it full screen.

To try it on your own repositories, download Safeguard and follow the CLI documentation.

Quick check

Sources

• Google Patents, US3338696A, Sheet forming apparatus (Stuart M. Dockerty)
• Google Patents, US8210001B2, Method of producing uniform light transmission fusion drawn glass (Corning)
• Corning, Corning to Invest in Gen 10.5 LCD Glass Substrate Facility in Hefei (1 December 2015)
• Sharp, Operations Begin at World's First LCD Panel Plant Employing 10th-Generation Glass Substrates (1 October 2009)
• Sharp, IGZO technology page
• Nippon Electric Glass, flat panel display glass products
• US Department of Energy, OLED Basics
• Google Patents, US9188856B2, Type of fine metal mask used in OLED production and the method of manufacturing it
• IEC, IEC 62341-1-1:2009, OLED displays, Part 1-1: Generic specifications
• IEEE Spectrum, Digging Into the New QD-OLED TVs
• Apple Newsroom, Apple unveils Apple Watch Series 5 (September 2019)
• Apple Newsroom, Apple unveils iPhone 13 Pro and iPhone 13 Pro Max (September 2021)
• Apple, iPad Pro technical specifications
• VESA, VESA Standards (free standards list including DDC/CI, DisplayID, E-EDID, MCCS)
• Microsoft Learn, Using an INF file to override EDIDs
• Linux kernel documentation, EDID (admin guide)
• Linux kernel documentation, Mode setting helper functions and EDID helper reference
• Linux kernel documentation, Kernel mode setting
• ddcutil documentation, introduction
• ddcutil documentation, frequently asked questions
• ddcutil documentation, command overview
• fwupd, Plugin: Mediatek Display Controller
• fwupd, Plugin: Synaptics MST
• LVFS documentation, Introduction

Last reviewed: 23 October 2026.