Installation: most faults are not faults
Of the units sent back to us as dead, roughly nine in ten work when they arrive. They were fitted into a housing with no clearance, plugged into a port that browns out at 400 mA, or grounded through an anodised coating. This page is the checklist we run through on the phone — printed here so you can run it without calling.
Eight steps, and the numbers that matter
Tools: PH1 screwdriver, 0.5 N·m torque driver if you have one, 70 % isopropyl, lint-free cloth, multimeter, cable ties.
Dry-fit before you drill anything
Set the frame on the housing with no screws and press each corner. It must not rock. A 0.3 mm rock at one corner becomes a 2 mm positional error at the opposite one, and calibration cannot remove it because the frame is physically twisted.
Clean the four faces the clamps touch
70 % isopropyl on a lint-free cloth. A single dust particle under a clamp lifts the frame by roughly 0.1 mm and shows up as a dead spot near that corner. Do not use window cleaner — the surfactant leaves a film that attracts dust.
Measure the lit area, not the diagonal on the box
A 55 inch display is 55 inch across the glass, but the lit area is smaller and the bezel is 8–20 mm. The frame is cut to clear the lit area. If you give us the diagonal instead, expect to shim.
Fix opposite corners first, then snug — never tight
M3 into aluminium: 0.5 N·m, and 0.6 N·m is already the point where the thread starts to pull. Two screws, opposite corners, then the other two, then a final pass at torque. A stripped insert in a 2.5 m extrusion is not repairable on site.
Check the glass sits inside the active area
Glass up to 5 mm is standard; 8 mm needs the wide-gap version. What matters more than thickness is the edge: if the edge of the cover glass crosses the beam path, you get a permanent dead line down that side.
Route the USB cable away from the backlight driver
The supplied cable is 1.8 m. USB 2.0 is reliable to about 3 m passive; beyond that use a powered hub or an active extension, and never a passive 5 m cable with a 40-point frame. Keep the run away from the inverter or LED driver — a 200 Hz report rate is easy to disturb.
Ground the frame to the chassis
Measured resistance under 4 Ω. Anodising is an insulator, so a screw through the coating is not a ground: use a toothed washer or scrape a 5 mm spot under the lug. An ungrounded frame in dry air ghosts, and the fault looks like a firmware bug.
Power on, then calibrate before you judge it
Plug in, let the OS enumerate the digitizer, then run calibration with the frame already fixed in its final position. Calibrating on the bench and then mounting the frame throws the mapping out by the thickness of your brackets.
Clearance table
| Dimension | What to leave | Note |
|---|---|---|
| Frame outer size | See the size page for your diagonal | ± 1.0 mm on the drawn dimension |
| Housing cut-out | Frame outer + 4 mm on width and height | 3 mm extra at the cable exit |
| Frame depth | 9.5 mm | Add 2 mm if the USB plug is moulded, not overmoulded |
| Glass thickness | ≤ 5 mm standard | 8 mm with the wide-gap frame |
| Front clearance | 0 mm — frame sits proud of the glass | Bezel visible from the front, by design |
| Screw | M3 × 6, four corners, 0.5 N·m | Inserts in the extrusion, not in the housing |
Six steps, and where the yield goes
A capacitive panel is a bonding job before it is an electronics job. Most of the failure modes are decided before the controller is ever powered.
Decide the bonding before you cut the housing
OCA sheet, liquid OCR, double-sided tape, or an air gap. OCA is the cleanest and costs the most. Tape is fine indoors and fails outdoors when the gap fogs. The stack-up decides your housing depth, so this is a drawing-stage decision, not an assembly-stage one.
Bond in a clean area, not on the line
One particle between sensor and glass is a bubble, and a bubble over 0.3 mm in the active area is a reject. Laminating in the same room as the frame saw is the single most common cause of a bad first batch.
Keep the FPC bend radius at three times its thickness
No creases, no folds at 90°, no tension on the connector. Tape the tail down within 10 mm of the connector so vibration cannot walk it out. A tail that fails in the field always fails at the crease.
Ground the sensor shield to the chassis
Same rule as infrared: under 4 Ω. A floating shield picks up switching noise from the power supply and shows up as a jittering cursor that only appears on some units.
Match the cover glass to the firmware
1.1 mm, 1.8 mm and 3 mm glass need different sensitivity settings. The panel is tuned for the stack you tell us about; if you change glass after approval, the tuning has to change with it.
Test twenty points and the four edges
Twenty simultaneous fingers, then a finger dragged along each edge 5 mm in from the border. Edge linearity is where a panel that passed a five-point centre test still fails in the field.
The five-minute check
- Frame or panel sits flat, no rock at any corner, no gap you can see light through.
- All screws at 0.5 N·m, none missing, none over-torqued.
- Frame-to-chassis resistance under 4 Ω, measured, not assumed.
- USB cable has strain relief and does not pull on the controller connector.
- Nothing is touching the active area — no tape, no protective film left on, no glass edge in the beam path.
- Host powered from its real supply, not a bench USB charger.
Then calibrate. Judging a frame before calibration is judging a scale before it is zeroed.
Symptom, cause, fix
Ten faults that cover most of what we see. If yours is not here, the five-item request on the support page gets you a diagnosis in one reply.
| Symptom | Likely cause | What to do |
|---|---|---|
| Cursor sits a few millimetres off, worse toward one corner | Frame mounted after calibration, or a twisted frame | Re-run calibration with the frame fixed; check for rock at the corners |
| Ghost touches in dry weather | Frame not grounded, or ground through anodising | Toothed washer to bare metal, measure under 4 Ω |
| Works, then drops out when the display wakes | USB port browning out — the frame peaks around 320 mA at 40-point | Powered hub, or a port that holds 500 mA; check with a USB power meter |
| Dead vertical stripe | One LED tube failed, or a glass edge crossing the beam | Swap the tube; move the glass inside the active area |
| Touches register only near the centre | Dirty filter strip, or calibration done with an object on the glass | Clean the strip with isopropyl, remove everything from the surface, recalibrate |
| Two-finger gestures report as one finger | Controller flashed in single-touch mode | Send us the label photo — we ship a reflash tool or swap the controller |
| OS shows “HID-compliant mouse” | Wrong HID descriptor | Same fix: reflash to the touch descriptor |
| Capacitive panel jitters only on some units | Sensor shield not grounded, or noise from the SMPS | Bond the shield to chassis, route the tail away from the supply |
| Capacitive dead band along one edge | Air gap too large, or printed border overlapping the sensor | Reduce the gap; check the border width against the drawing |
| No touch on Linux, fine on Windows | hid_multitouch not loaded | modprobe hid_multitouch, then check dmesg |
We can watch the first one go in
For any frame above 86 inch, any curved or LED-wall build, and any first capacitive bonding run, we join by video call. You prop a phone against the housing; our engineer watches clearances, screw order and cable routing while you work, and stops you before the mistake rather than after it. No charge, and it usually saves the cost of the call in the first ten minutes.
Installation questions
How long does a first installation take?
Can I install the frame without removing the display from the wall?
What USB cable length is safe?
Do I need an earth connection for an indoor kiosk?
Can you join the installation remotely?
The panel is bonded and there is a bubble. Can it be saved?
Send a photo of the housing.
One photo of the opening and the display's lit area is enough for us to confirm clearance, frame size and the cable route before you cut anything.