October 13, 2022
Driver ICs on small LCDs are attached in two mainstream ways. The difference runs all the way through to thickness, vibration tolerance, humidity tolerance, assembly and repairability.
Small LCD driver ICs are attached in two mainstream ways. COG (chip-on-glass) bonds the IC directly to the edge of the glass, so the module is very thin and the run from IC to glass traces is short, allowing very fine traces. COB (chip-on-board) places the IC on the PCB, wire-bonded and then encapsulated with epoxy, structurally closer to a conventional board.
That difference propagates through everything downstream: thickness, vibration tolerance, humidity tolerance, assembly and repairability all fork here. COG and COB are not advanced versus obsolete; they are two different technical routes, each suited to different operating conditions. Choose the wrong route and no amount of process tuning recovers it.
COG's most visible advantage is thinness. Modules can go below two millimetres, the glass edge outputs directly, interconnect runs are short and electromagnetic interference is low. 1602 character formats and 122x32 to 240x128 graphic formats are all mature, with a wide choice of glass and drivers. Any product sensitive to thickness and weight — ultra-thin remote controls, wearables, compact panels — has COG as the default answer.
Its boundaries should be stated clearly. The IC bonds directly to the glass, so glass-edge stress, bending during assembly and shock during transport can all affect reliability, and the assembly process demands more precision. The seal border is sensitive to moisture, so reliability in long-term high-humidity environments needs dedicated design. In short, COG is friendly to thin and light; it is less friendly to rough service.
COB is the opposite. The driver IC sits on the PCB, wire-bonded and then epoxy-encapsulated as a whole, giving clearly better vibration and humidity tolerance, high assembly tolerance, and board-level repair when something fails. The supply chain is mature and cost is usually lower. 1602, 2004 up to 320x240 character and graphic formats are classic solutions, used for over a decade in industrial instruments and appliances.
COB's boundary is volume. Putting the IC on the PCB means the module must leave room for it, so both thickness and area rise, and squeezing it into a two or three millimetre space is essentially impossible. The inductance of bond wires is also less favourable than COG's direct bond under high-frequency driving, making it sub-optimal for high-resolution, high-refresh designs. Thinness and operating conditions — the trade-off sits in that exchange.
Practical selection comes down to three questions. First: where does this screen go, and how many millimetres are available for it — the answer decides whether COG is possible. Second: what is the operating environment, will it vibrate, how humid, will it be repaired — the answer decides whether COB is needed. Third: cost and capacity — both are mature, so this is rarely decisive.
With those three answered, the route is usually clear. The non-standard remainder can be developed per order: non-standard inch sizes and resolutions can be developed, and interface, logic voltage, viewing direction, FPC length and pin definitions can all be adjusted to the design. We have mature COG and COB solutions alike; submit the content, dimensions and operating conditions and a solution and quotation come back within twenty-four hours. Selection is not picking a product; it is choosing a route that matches your device.
The COG versus COB argument is usually framed as technology choice, but for small devices the more consequential variable is how the module connects to the host board. COB modules with pin headers are forgiving: they can be socketed, replaced in the field, and handsoldered on a short production run. The same module in a zebra-strip or flexible tail configuration saves space but makes field service impractical and puts alignment requirements on the assembly process. On a few hundred units, a header-based module with a slightly larger footprint is often the lower-risk choice.
Zebra strip assemblies in particular deserve a warning. They depend on consistent pressure and clean surfaces, and a housing that flexes under temperature will lose contact intermittently. If a zebra design is unavoidable, specify the compression range and test a fully assembled unit through temperature cycling rather than the bare module. We state the connection method and its mechanical requirements in the quotation so the assembly risk is visible before the order is placed.
Most COG versus COB debate stops at performance comparison. The question that actually decides it in the field is usually skipped: what happens when this screen fails. COG bonds the driver IC to the glass, so a failure means replacing the whole module, with no possibility of field repair. COB keeps the driver on the PCB, so there is theoretical rework space even though in practice replacement is more common.
Product positioning decides the answer completely. For a disposable consumer device with extreme cost sensitivity, the thinness and saving of COG is correct and repair is irrelevant. For equipment entering an industrial site, expected to serve for years and to be replaced in the field, the choice should tilt toward maintainability. This is not a technical ranking, it is a consequence of product type.
Small-batch projects add one more consideration: failed units have to be analysed during the trial. If the module cannot be opened, failure analysis stops at replace it and no root cause is obtained. We flag this trade-off at customisation stage so the client knows at quotation what has been chosen, rather than discovering there is no way back after the failure.
Once COG is chosen, the customisation happens on four layers, and it helps to know which ones are cheap. The glass outline and the segment layout sit on the mask, so they are tooling: a COG LCD module custom project pays for them once and then reuses them for every unit. The driver IC and the bias scheme are a choice from a fixed menu. The FPC is artwork, cheap to revise at prototype stage and expensive after tooling. The bezel is the cheapest of all and the one buyers change most often.
A monochrome graphic LCD in a custom size is a different animal from a segment panel. Here the mask defines a pixel matrix — 128x64, 122x32, 160x128 — and the customisation is the outline around it plus the controller that drives it. The trade-off is that a graphic panel can show any content the firmware draws, at the cost of a controller and a font table, while a segment panel is cheaper per unit and far cheaper to drive. Products that show the same icons for a decade should be segment; products whose screen content will change with firmware should be graphic.
Tell us your project scenario and requirements, and we will match the right solution and provide a quote within 24 hours.
Get a Free QuoteCOG and COB character and graphic LCDs, PMOLED/AMOLED microdisplays, 0.69–10.1 inch TFT colour panels and LED backlight modules across six product families and 61 standard models. Non-standard sizes, resolutions, interfaces, viewing directions and FPC layouts developed to drawing; standard, industrial and wide-temperature grades; supplied to RoHS, REACH and CE with full compliance documentation and a 12-month warranty that can be extended per project.
Explore the Display Module Solution →