Mechanical, Divider Valve or Servo Closed Loop: Choosing Between Three Ways to Sync Cylinders

A specification for a machine that needs several cylinders moving together usually offers three options: add a synchronising shaft or rack and pinion, use a divider or synchronising valve, or fit every cylinder with its own valve and magnetostrictive displacement sensor and close a loop. The prices differ by a multiple, and the maintenance requirements differ just as much.

It is easy to be led by price. Try a different thread instead: the three methods differ in what force actually enforces the synchronisation. Once that is clear, the boundaries fall out.

Mechanical sync: structure enforces it, not control

Rack and pinion with a torque shaft, a rigid linkage, a shared synchronising shaft — what these share is that the two cylinders are hard-coupled. Neither can go where the other is not. Sync is not controlled, it is connected.

The advantages are concrete: no dependence on electrics or oil condition, sync maintained through power loss, and a failure mode anybody can read — either the gearing wears into backlash or a coupling works loose. Maintenance is mechanical work, handled by a millwright.

The costs sit in three other places. It takes space, and a long span needs a stiff shaft, which means a thick and heavy one. Inertia is high, which hurts fast cycling. Wear produces backlash, and that error grows rather than returning to zero. The wider the span and the longer the stroke, the harder the structure becomes.

Divider valve sync: splitting flow

Divider valves, synchronising valves and gear flow dividers all belong here: one supply is split in a fixed ratio and each path drives its own cylinder. There is no position value in the circuit; the basis for splitting is flow.

The advantages are simplicity, low cost and no electrical dependency. The drawback is equally clear: split tolerance, asymmetric load, oil compressibility, internal leakage and oil temperature drift stay uncorrected, so the position difference accumulates along the stroke. When a datasheet quotes a split accuracy percentage, check what flow, what pressure differential and what oil temperature it was measured at.

Where it fits shares one feature: the error never gets a chance to accumulate into the product. Short strokes, intermittent motions, and a mechanical stop taking over at the end of travel — those conditions make a divider valve economical.

Servo closed loop: position feedback maintains it

Each cylinder gets a proportional or servo valve and a magnetostrictive displacement sensor, plus a controller that compares actual positions and corrects each cycle. The synchronisation is maintained by the control loop rather than by the valve characteristic.

The accuracy band is genuinely different, but the costs need stating: valve and magnetostrictive displacement sensor hardware, sensitivity to oil cleanliness, commissioning effort, and the need for somebody on site who can read the controller parameters. One more cost is easy to miss — sync accuracy becomes tied to the position feedback itself. A reading problem becomes a sync problem, so mounting and calibration have to be done properly. Mounting is covered in external and in-cylinder mounting, calibration in zeroing with a mechanical datum.

Side by side

Item	Mechanical sync	Divider valve sync	Servo closed loop
What maintains sync	Rigid structural coupling	Flow split in a fixed ratio	Position feedback correction
Does the error accumulate	No, but backlash grows	Yes, along the stroke	No, corrected continuously
Response to load change	Structure absorbs it, may distort or bind	Poor, load difference becomes position difference	Good, loop compensates
Long spans and strokes	Structure becomes impractical	Largely unaffected	Largely unaffected
Electrical dependency	None	None	Strong
Maintenance skill	Mechanical only	Mechanical, mostly hydraulic	Hydraulic plus controls
Diagnosability	Inspect backlash and wear	Effectively invisible until it fails	Deviation and trend readable online
Cost band	Depends on structure, rises fast with span	Low	High

Three steps to choose

• Step one: convert the sync requirement into a percentage of stroke. Around one percent allows a conversation about divider valves. A few parts in a thousand, or a fixed figure in millimetres, means closed loop. The conversion is covered in building an error budget.
• Step two: ask whether the error gets a chance to accumulate. A motion of a few tens of millimetres that stops at a hard stop gives the error no time. A motion that runs for minutes, or that must hold sync at any intermediate position, accumulates every time.
• Step three: check whether mechanical guidance is already doing part of the work. Tie rods on a four-column press, guideways on a gantry, a scissor linkage — each of them carries some of the synchronising constraint, which allows a lighter electrical solution.

Those three steps settle most cases. The genuinely awkward ones are the middle ground where none of the three fits cleanly, and that is where combinations come in.

Where the methods are combined

The three approaches can share one machine. Two combinations are common.

Mechanical guidance, hydraulic drive, position monitoring. The guideways or tie rods constrain the sync, the hydraulics provide the force, and the magnetostrictive displacement sensor readings are used to detect abnormality. In this arrangement it only monitors and takes no part in the loop, so cost comes down — accept that it can only report a problem, not correct it.

Master-slave. One cylinder follows the command, the other follows the first cylinder's actual position. The two cannot drift apart independently, at the price of the master's following error being passed straight to the slave, capping overall accuracy at the master. A fully peer-to-peer arrangement costs more but does not pass error in one direction only.

Where redundant monitoring is specified there is one further consideration: two magnetostrictive displacement sensors only earn their cost if the readings are genuinely compared. Installation without comparison achieves nothing — the method is in double sensors that are never compared.