Magnetic-Levitation Motion on a UV Flatbed Printer, and Why Carriage Drive Shows Up in Your Print

The complaint usually sounds the same. A large flat panel comes off the bed with faint horizontal stripes running across it, visible when you tilt it to the light and invisible in a photograph. The operator blames a nozzle. The supplier asks for the file. Somebody suggests more passes, which makes the stripes fainter and the job slower, and everyone agrees to move on.

Banding is a motion problem at least as often as it is an ink problem, which is why a magnetic levitation motion system on a UV flatbed printer is worth understanding before you sign for one rather than after. The carriage drive is the part of the machine that decides where each drop lands in the scan direction. If it delivers velocity that is not constant, or position feedback that is not clean, no amount of head quality will save the print.

Three ways to move a print carriage

Every flatbed moves a head carriage along a scan axis and steps the gantry or the bed along a second axis. Three drive technologies dominate, and they fail differently.

The honest summary: belt drive is cheap and fine at moderate speed on short axes, and it is where visible banding most often originates on low-cost wide-format machines. Ball screws are accurate but do not like being long and fast at the same time, which is a problem on a 2.5 m scan axis. Linear motors were adopted in this class of machine because they solve the long-and-fast case, not because they are exotic.

What "magnetic levitation" is actually describing

In printer marketing the term covers a range of engineering. At the drive level it means a linear synchronous motor: magnets along the beam, coils on the carriage, and force applied directly with no belt, screw or gear in between. There is nothing to stretch, nothing to backlash, and nothing to wear in the force path.

The word "levitation" refers to how the carriage is supported, and that is a separate question from how it is driven. A carriage can be driven by a linear motor while still riding on conventional recirculating-ball linear guides. It can also be supported without mechanical contact by magnetic preload or air bearings. Both are real designs, and both are sold under the same word.

So make it a question rather than an assumption, and ask it in two parts:

1. Is the drive a linear motor — magnets and coils, no belt or screw?
2. What supports the carriage — recirculating-ball linear guides, or a genuinely non-contact bearing?

The answers change your maintenance schedule and your spares list, and neither answer is a bad one. What you are buying is a specification, and the point of a specification is that it is written down. Longrun's published listing for its 2513 flatbed states a magnetic-levitation motion system carrying Ricoh Gen-6 printheads over a 2500 x 1300 mm bed; the guideway type, the repeatability figure and the lubrication interval are not published, so they belong in your first technical email.

The encoder decides your print, not the motor

Here is the detail that separates people who have run these machines from people who have read the brochure.

A linear motor does not know where it is. It is a force actuator in a closed loop, and the loop is closed by a linear encoder — a scale running the length of the axis and a read head on the carriage. The controller reads position from that scale, and the printhead fires its nozzles against those encoder counts. Drop timing is derived from encoder position, not from a clock.

That has three consequences.

Encoder resolution sets your placement floor. At 1,200 dpi addressable resolution, one pixel is about 21 µm. At 600 dpi it is about 42 µm. An encoder that resolves far coarser than the pixel you are trying to place cannot place it, however good the motor is. Ask for the encoder type — glass scale or magnetic tape — and its resolution in µm per count.

A dirty encoder produces defects that look like head failure. UV printers throw a fine ink mist that settles on everything inside the gantry, and the encoder scale is inside the gantry. Contamination on the scale gives the read head a bad count, the controller fires early or late, and you get a vertical streak or a short misregistered band that moves around between prints. Operators who know the machine clean the encoder strip on a schedule with the manufacturer-specified solvent. Operators who do not, replace healthy printheads. Find out where the strip sits, how it is accessed, and what it is cleaned with, and put it in the commissioning and operator training plan before the trainer flies home.

Bidirectional printing doubles your exposure to timing error. Printing on both the left-to-right and right-to-left strokes nearly doubles throughput. It also means any lag between the encoder count and the actual firing shows up as a mismatch between alternate passes — a sawtooth on vertical edges, or a fine two-line beat across a solid. The correction is a bidirectional offset calibration, and it drifts with carriage speed, head height and temperature. A shop that cannot hold bidirectional registration ends up printing unidirectional and losing half the speed it paid for. Ask how the bidirectional offset is calibrated, whether it is stored per print mode and per head height, and how long the calibration takes.

Banding: read the pitch to find the cause

Banding has a spatial period, and the period tells you where to look. This is the fastest diagnostic in the shop and it costs nothing.

• Bands at the swathe pitch — the width the gantry steps between passes — point at the step axis or at the nozzles at the edge of the swathe. This is the most common banding of all, and it is a Y-axis and nozzle problem, not a carriage problem.
• Fine, regular bands at a much shorter period point at velocity ripple in the scan axis or at the encoder. Iron-core linear motors produce a small periodic force variation as the coils pass the magnet pitch; ironless designs largely avoid it. If the band pitch matches a mechanical or magnetic pitch on the axis, you have found it.
• Bands that change with print direction are a bidirectional offset problem.
• Bands that appear only after an hour of running are thermal, and are covered below.
• Irregular, non-periodic mottling is usually substrate, vacuum hold-down or ink, not motion.

The step axis is the one buyers never ask about

Almost every specification conversation is about the scan carriage, because that is the axis with the impressive drive technology on it. But the swathe-to-swathe registration — the accuracy with which the gantry steps a few millimetres or a few centimetres between passes — is where most visible banding is born. A scan axis that is accurate to a few microns paired with a step axis driven by a stepper motor and a belt will still band. Ask which drive technology is on the Y axis, and what its repeatability is, in the same email in which you ask about the X.

Heat, and why the fourth hour differs from the first

Linear motors put current through coils, and coils get warm. That heat goes into the carriage and the gantry beam. Add the thermal load from whatever cures the ink, and a printer gantry that has run for an hour is not the same length as one that has just been switched on.

The arithmetic is unforgiving on a long axis. Aluminium has a coefficient of thermal expansion of roughly 23 µm per metre per °C; steel is roughly half that. A 2.5 m aluminium beam that warms by 5 °C over a shift grows about 290 µm — around fourteen pixels at 1,200 dpi. In practice much of that is common-mode and gets absorbed by the encoder, if the encoder scale is mounted so that it expands with the structure it measures. If it is not, the drift lands in your print.

You do not need to solve that as a buyer. You need to know it exists, and ask two things: is there a specified warm-up routine before production printing, and does the machine hold registration on a long continuous run. Then test it. A useful acceptance test is a two-hour continuous run of the same file, with the first sheet and the last sheet measured against each other on a registration cross at all four corners. That is a better use of a pre-shipment test print than a pretty photograph, and it slots naturally into a wider factory acceptance test for machinery bought from China. Longrun lists a test-print stage in which the buyer's own files and substrates are run before shipment — that is the moment to spend it on a two-hour drift test rather than a single showpiece panel.

The drop is in the air while the carriage moves

One more piece of physics, because it explains why motion accuracy stops being the limiting factor on some jobs.

A drop leaves the nozzle and crosses the head-to-substrate gap while the carriage is still moving. Its sideways landing position is shifted by roughly the gap multiplied by the carriage speed divided by the drop velocity. With a 1.5 mm gap, a carriage at 1 m/s and a drop velocity around 7 m/s, that is a lateral shift of about 210 µm — ten pixels at 1,200 dpi. Machines correct for it with a fixed timing offset, and it disappears.

What does not disappear is the variation. Change the gap by 0.5 mm — a warped sheet, an uneven part, a thicker substrate with the head height not reset — and the shift changes by around 70 µm, roughly three pixels, with no error at all in the motion system. That is why gap control and substrate flatness deserve as much attention as carriage specification, and why the choice of printhead family matters on uneven work: heavier drops from larger-drop heads cross the gap faster and are less sensitive to it, which is the trade covered in choosing between Ricoh Gen-6 and Seiko heads. Holding the sheet dead flat, in turn, is a vacuum-table question, covered in what a 2500 x 1300 mm bed actually holds.

Common questions

Will a maglev carriage stop my machine banding?

Not on its own. It removes belt stretch, backlash and reversal error from the scan axis, which eliminates one family of banding. It does nothing about step-axis error, nozzle dropouts, substrate lift, ink issues or bidirectional calibration drift. Treat it as one solved problem out of several, and diagnose the rest by band pitch.

Is a linear-motor axis harder to maintain?

The drive has no contacting parts, so there is nothing in the force path to wear, lubricate or tension. The maintenance moves elsewhere: keeping the encoder scale clean, keeping the guideway serviced according to its type, and keeping cable chains healthy on an axis that accelerates hard. Ask for the maintenance schedule as a document, with intervals and consumable part numbers, before shipment.

Can the carriage be moved by hand when the machine is off?

On a linear-motor axis, usually yes — there is no self-locking screw or gearbox holding position, so the axis is back-drivable when unpowered. That is convenient for service and a hazard if someone pushes the carriage into a raised part. It is worth covering explicitly in operator training rather than discovered by a new hire.

How do I compare two machines on motion accuracy?

Ask both for positioning repeatability in µm, maximum carriage speed and acceleration, encoder type and resolution, and the drive technology on both axes. Then ignore the numbers and run the same file on both, in the same print mode, on the same substrate, and look at the result at a low angle under a single light source. Specifications tell you what to ask about; a test print tells you what you are buying.

What to ask the supplier next

Eight questions to send with the drawing and substrate list. Tie the answers to the acceptance test, and tie the acceptance test to your payment schedule rather than to a delivery date — the logic behind milestone payments on industrial equipment from China, and the same comparison discipline the general UV flatbed printer buyer's guide applies elsewhere in the specification.

1. Drive technology on the X (scan) axis and on the Y (step) axis — stated separately, named as belt, ball screw or linear motor.
2. Carriage support: recirculating-ball linear guides or non-contact bearing, and which surfaces need lubrication.
3. Positioning repeatability in µm, and how it was measured — the measurement method matters as much as the number.
4. Maximum carriage speed and acceleration, and the speed actually used in the production print mode you will run.
5. Encoder type and resolution in µm per count, where the scale is mounted, how it is accessed and what cleans it.
6. Bidirectional calibration: how it is performed, whether it is stored per print mode and per head height, and how long it takes an operator.
7. Warm-up and thermal behaviour: is there a specified warm-up, and will the supplier run a two-hour continuous drift test with your file before shipment?
8. Maintenance schedule as a document, with intervals, part numbers and the spares to hold on site, agreed before the machine is crated.

Longrun Printing Machinery is a practical example to put those questions to: a Shenzhen manufacturer established in 2009 in Longgang district, with a 40-person engineering and assembly team producing roughly 300 machines a year, whose 2513 flatbed is listed with Ricoh Gen-6 printheads on a magnetic-levitation motion system over a 2500 x 1300 mm bed, sold from a single unit at 0.5–1.5 t depending on configuration, with a pre-shipment test-print stage and remote and on-site setup, training and spare parts listed in its process. The repeatability figure, the encoder specification and the maintenance interval are not published numbers — so ask for them in writing, and make the two-hour drift test part of what the balance payment is released against.