OLED LTPO Display Technology: How Variable Refresh Rate Works

Low-Temperature Polycrystalline Oxide (LTPO) does not change how an OLED emits light. It changes how each pixel is held and refreshed — combining two different thin-film transistor (TFT) technologies on the same backplane to enable truly variable refresh without flicker or excess leakage.

1. Why OLED Needed a New Backplane

An OLED pixel is self-emissive. The organic stack emits light proportional to current, and that current is controlled by a TFT backplane behind every subpixel. The organic materials get the attention, but the backplane dictates efficiency, uniformity, and how often frames must be rewritten.

Traditional smartphone OLEDs used LTPS (Low-Temperature Polysilicon) backplanes. LTPS is excellent for driving high current through a small transistor — essential for high brightness and high pixel density. Each LTPS TFT has very high electron mobility (80–100 cm²/V·s) and can switch quickly, which is ideal for a 120Hz scan.

The problem: LTPS leaks. The switching TFTs that hold a pixel's voltage between refreshes cannot hold charge for long. Even when the image is static — an Always-On Display clock, an e-book page — an LTPS panel must still be refreshed at 60Hz or higher, otherwise the stored voltage droops, luminance drops, and the panel flickers. That mandatory high refresh wastes power on the two largest consumers in the display system: the display driver IC (DDIC) and the GPU/display pipeline.

The industry tried to compensate with tricks. Panel Self Refresh (PSR) and frame buffering inside the DDIC can avoid GPU wake-ups, but the panel itself still had to be driven. What was needed was a backplane that could both drive hard and hold.

2. LTPS vs. Oxide TFTs: A Tradeoff

IGZO (Indium Gallium Zinc Oxide) is the alternative. IGZO is an oxide semiconductor with electron mobility an order of magnitude lower than LTPS, around 10–20 cm²/V·s, so it cannot source as much current in the same area. But its off-state leakage is extraordinary — nearly 1000x lower than LTPS. An IGZO TFT can hold a pixel voltage for well over a second without significant droop.

That is why oxide backplanes dominate large-format displays. An iPad Pro or a notebook LCD using IGZO can drop to 10Hz or 30Hz easily. Smartphones could not adopt pure IGZO because at 400–500 PPI, the larger IGZO driving transistors would force lower aperture ratio, lower brightness, or larger bezels.

The insight behind LTPO was to stop choosing.

3. LTPO: Two Transistors, One Backplane

Apple coined the term LTPO in 2018 with the Watch Series 4, though the underlying hybrid oxide concept was co-developed with LG Display. The implementation is deceptively elegant: use LTPS where you need speed and current, and IGZO where you need to hold.

In a typical LTPO 7T1C (seven transistors, one capacitor) pixel, T1 — the drive transistor that sets OLED current — remains LTPS. The emission control and data-write switches (T2, T4-T6) are also LTPS to maintain fast row scan at 120Hz. The critical change is T3 and T7, the transistors that connect the gate node of T1 to the storage capacitor. These become IGZO.

Why that node? The gate voltage on T1 (Vgs) is the analog memory of the pixel. If that node leaks, luminance shifts. By guarding the single most sensitive node with oxide transistors, LTPO allows the entire pixel to hold its programmed luminance for tens or even hundreds of milliseconds without leaking. The DDIC can then skip refresh cycles entirely — no scan pulses, no data reprogramming, no toggling.

Supporting this requires two independent gate driver chains: a p-type LTPS gate driver (pScan) and an n-type oxide gate driver (nScan), with separate voltage rails. The DDIC sequences them: IGZO gates are pulsed only when the pixel needs reprogramming; otherwise they remain off, sealing the charge on Cst. This dual-gate architecture is why LTPO adds mask steps and cost — typically 2–3 extra masks versus LTPS — but the yield hit has shrunk dramatically since 2022 as oxide processes matured.

For readers tracking foldable display durability, this same backplane consideration matters even more: foldable OLEDs now use LTPO to keep power down when the cover display is in AOD mode, despite the larger panel area.

4. How 1–120Hz Variable Refresh Actually Works

LTPO itself does not magically vary refresh. It enables it by removing the lower bound. The adaptive behavior is orchestrated by three layers working together:

The DDIC and Timing Controller

Modern LTPO DDICs (Samsung S6E3HC5, Novatek NT37801) integrate a frame-rate controller with an internal GRAM capable of storing a full frame. When the Application Processor (AP) signals that content is static — via Android's ASurfaceTransaction_setFrameRate or iOS's CADisplayLink.preferredFrameRateRange — the DDIC can enter a self-refresh mode. It continues to hold the last frame on-panel while telling the AP and GPU to enter low-power states.

The scan itself becomes intermittent. At 120Hz the DDIC drives every 8.33ms. At 10Hz it drives every 100ms. At 1Hz, once per second. Between drives, the scan drivers are power-gated and the source channels are tri-stated. The oxide TFTs ensure the pixel does not decay visibly in between.

The Operating System Compositor

On Android 14+, SurfaceFlinger and the Android 16 architecture changes to the display pipeline allow per-window refresh hints. A video window can request 24Hz, the status bar can sit at 60Hz, and a scrolling list can burst to 120Hz — the DDIC picks the lowest common denominator or uses panel-level region refresh where supported. iOS does similar with ProMotion, dynamically classifying content as idle, video, or interactive.

This is not free. Abrupt refresh changes expose OLED gamma shift: at very low refresh, the long hold time makes the human eye sensitive to tiny leakage-induced luminance drops (often visible as 1Hz flicker in dark grey). Vendors counter this with multi-gamma tables and adaptive emission duty control — shortening OLED on-time at low refresh to keep perceived luminance stable.

// Let the display pipeline know this Surface prefers low refresh for static content
val transaction = SurfaceControl.Transaction()
transaction.setFrameRate(
    surfaceControl,
    10.0f,  // preferred refresh, DDIC may quantize to 10 / 30 / 60
    SurfaceControl.FRAME_RATE_COMPATIBILITY_DEFAULT,
    SurfaceControl.CHANGE_FRAME_RATE_ONLY_IF_SEAMLESS
)
transaction.apply()

// Compose modifier for 120Hz during gesture, else idle at 10Hz
Modifier.preferredDisplayMode(
    refreshRateRange = 10f..120f
)

5. LTPO 2.0 vs. LTPO 3.0: What Changed?

First-gen LTPO (2019–2021) switched between fixed steps: 10/30/60/120Hz. The quantization was coarse because the DDIC could only use a few pre-calibrated gamma curves. In low-brightness scenes, stepping from 60Hz to 30Hz produced a visible luminance bump.

LTPO 2.0 (2022–2024) introduced granular frequency control and adaptive gate pulse modulation. Samsung Display's Eco² OLED Plus and LG's LPTO+TSP variants widened the range to 1–120Hz with 1Hz granularity, added LTPO-aware demura compensation to correct oxide TFT non-uniformity, and decoupled the touch sampling rate from the display refresh — letting touch run at 240Hz while the panel refreshes at 1Hz for AOD.

LTPO 3.0, now shipping in 2025–2026 flagships, pushes two key advances:

1. Oxide drive participation. Some vendors now use IGZO for a subset of drive duties at low refresh, further cutting leakage paths. BOE's LTPO+ design replaces one additional LTPS emission TFT with a stabilized top-gate IGZO, reducing the refresh lower bound drift from ±0.8 nits to ±0.2 nits at 1Hz.

2. Intra-frame refresh partitioning. Rather than refreshing the entire panel at the same cadence, the DDIC can refresh the region containing a scrolling animation at 120Hz while keeping the status bar at 30Hz. This is why power savings scale better than the simple frequency ratio suggests. For on-device ML workloads that overlay real-time inference results, region refresh avoids forcing the whole display to 120Hz for a small changing viewport.

6. Power Savings: Where the Milliwatts Go

On a 6.7-inch QHD+ OLED, the display subsystem accounts for 40–55% of system power at typical brightness. At 120Hz, the DDIC plus gate drivers consume roughly 180–220 mW; the OLED emission itself around 300–500 mW depending on APL (average picture level). Dropping to 10Hz does not cut emission power — the pixel is still on — but it collapses driver and interface power by 70–85%.

In instrumented tests, a looping static reading session (200 nits, light content) lasted 18% longer on an LTPO 3.0 device versus the same panel forced to 60Hz LTPS emulation. The saving is most pronounced in Always-On Display: an LTPS panel cannot stay in AOD without refreshing at 60Hz (~90 mW), while LTPO at 1Hz sits near 25 mW — enough to keep AOD on overnight without noticeable drain.

There is an often-overlooked second-order effect. When the DDIC enters self-refresh, the AP's display pipeline can gate its clocks and DDR bandwidth drops. For workloads already running on-device LLMs or camera previews, that freed memory bandwidth directly translates to thermal headroom for sustained inference.

Display Modes reported by DDIC (via MIPI DSI):
  Mode 0: 1264x2780 @ 1Hz   (AOD, 8-bit, duty 15%)
  Mode 1: 1264x2780 @ 10Hz  (Idle / Reading)
  Mode 2: 1264x2780 @ 30Hz  (Video 30p)
  Mode 3: 1264x2780 @ 60Hz  (Video 60p / Default)
  Mode 4: 1264x2780 @ 90Hz  (Gaming balanced)
  Mode 5: 1264x2780 @ 120Hz (Touch / Scroll / Gaming max)

Seamless switch: Modes 1–5 use same timing base (Htotal/Vtotal) — no mode-set needed.
Actual frequency = IDLE_CNT + TE signaling; quantized by TCON look-ahead.

For buyers, the practical checklist is short: confirm the panel is hardware LTPO (not "LTPO-like" software VRR that merely toggles 60/120Hz), check that AOD operates at 1Hz or 10Hz rather than 30Hz, and look for LTPO 2.0+ decoupled touch if you rely on Always-On or stylus input. True 1Hz operation requires both the hybrid backplane and a DDIC with GRAM — one without the other is just marketing.

LTPO's role will only grow. As foldables push larger areas and AR micro-OLEDs demand microwatt idle power, the same principle — drive hard when needed, hold without leaking the rest of the time — is now the baseline for OLED efficiency. LTPS is not dead; it remains the cost-effective choice for mid-tier panels. But for flagship mobile displays, variable refresh is no longer a luxury feature. It is the backplane.