Quick answer: A converter scaler sits between a source that outputs one fixed timing and a panel that accepts another. It reads the incoming frame, deinterlaces it if needed, then writes the panel native format. It must also describe its abilities upstream through EDID, or the source may never start. Latency is the third limit: a frame or two of buffering is harmless for film and unacceptable on an arcade cabinet or a live clinical display.
The source is a set top box, a camera or an arcade board with exactly one output format, and the panel is an installed display that accepts exactly one input format. Neither end will negotiate. What absorbs the mismatch is a scaler with a frame buffer, and its three weak points are the same on every model.
Two fixed ends and the box between them
A scaler has to do more than change a number. It must sample the source timing, detect the field or frame structure, store at least one frame, then generate a new timing with its own pixel clock. Every stage adds delay, and every stage can fail on an unusual input.
Why a mismatch cannot be fixed passively
Passive adapters only rearrange wires. A panel that expects a progressive frame at one rate cannot be talked into accepting an interlaced field at another, because the two differ in structure, not only in connector.
What the box needs to know about the panel
The panel native resolution and refresh are the target. Scaler firmware holds a short list of supported outputs, and the installer picks one. Choosing a rate the panel does not accept gives a blank screen even when the conversion works.
Deinterlacing is where the picture is decided
Interlaced material arrives as two fields captured at different instants. Combining them is a decision, not a copy. The chipset can average the two, keep one and interpolate, or compare them and use motion adaptive work only where movement is detected. Each choice trades detail against artefacts.
The field pair problem
If a moving object shifted between fields, averaging produces a double edge on every frame. Keeping a single field avoids the doubling but throws away half the vertical resolution, which is obvious on static text.
What an inexpensive chipset does
Low cost designs average the two fields and scale afterwards. That suits slow footage and is visibly wrong for sport, scrolling credits or a fighting game, where edges smear into a comb pattern no downstream setting can remove.
The EDID it must present upstream
EDID is the small data block a display sends to describe what it accepts. A scaler has to answer that conversation on behalf of the panel, and what it claims decides whether the source sends a picture at all.
Claiming too much or too little
Claim too much and the source picks a format the box cannot build. Claim too little and it refuses to start, or boots into a fallback that looks permanent. The reliable setting is the panel native timing plus one conservative fallback.
| What the EDID claims | Source behaviour | Panel receives | Failure you see |
|---|
| One native timing only | Falls back or refuses | The correct format | No signal from a picky source |
| Native plus a low fallback | Starts on the fallback | Boot screen, then upscaled | Soft picture until source resets |
| A long list of modes | Picks the highest | A format the box cannot build | Black screen after handshake |
| No EDID response | Assumes a default | Whatever the default was | Wrong aspect or doubled image |
Latency the room cannot accept
Buffering a frame is unavoidable, since the box must read one complete frame before writing another. At 60 Hz a single frame is roughly seventeen milliseconds, and a design buffering two or three adds proportionally more. For film that delay is invisible.
Interactive rooms are different. On an arcade cabinet the delay sits between the joystick and the sprite, so a player adapts by aiming early and the game stops feeling correct. In a clinical setting it sits between the probe and the image, and a lagging view of a moving hand is a safety question rather than a comfort one. Ask for the measured figure in frames and take the smallest available.
Practical scenarios
- Arcade board into a flat panel: prioritise a box with minimal buffering, and accept a slightly softer picture in exchange for response.
- Legacy camera feed on a wall display: choose motion adaptive deinterlacing, since averaged fields smear anything that moves.
- Source will not output at all: the EDID it reads lists formats the box cannot build, so limit the list to native plus one fallback.
- No picture after a swap: re-power the source so it re-reads EDID from the new box instead of using its cached choice.
FAQ
Is a scaler the same as a converter?
No. A converter changes the signal type, for example analogue to digital. A scaler changes resolution and timing. Many boxes do both, which is why one word gets used for the other.
Can a scaler improve a low quality source?
It can fill the panel and remove interlacing artefacts, but it cannot restore detail the source never carried. Expect a clean, soft picture rather than a sharp one.
Why does the image shift to one side?
The box samples the blanking interval at the wrong point, so the picture starts late. Some units expose a horizontal position control, and a few pixels of adjustment corrects it.
Next step
Check host and port behaviour in the USB adapter standards guide, plan the desk end with the docking station guide, and read the long distance USB guide before the run passes ten metres.
See the matching cables and adapters
