Belt balancers

Balancer or manifold?

A manifold is just a chain of splitters, one per consumer, each taking what it needs off the belt and passing the rest along. It only evens out once every machine's input buffer downstream is already full – until then the first machine in the chain gets priority. That's fine when every consumer is identical, the total supply is at least the total demand, and the whole flow fits on one belt; it costs the fewest possible parts.

A balancer splits and merges the flow so every output carries an equal share from the first item onward, no buffering required. It costs more splitters and mergers than a manifold, but it's the only correct choice when consumers must start in sync, when supply is tight and starving one machine would be worse than running all of them slightly under, or when the flow itself has to be spread across several belts because a single belt can't carry it.

The two building blocks

Every balancer is built from exactly two parts:

• Splitter () – one belt in, up to 3 belts out, split evenly.
• Merger () – up to 3 belts in, one belt out, combined.

Every edge in a balancer still has to fit on a real belt. See Belts & pipes for lifts, curves and build cost; the capacities that matter here are:

Tier	Belt	Items/min
Mk.1	Conveyor Belt Mk.1	60
Mk.2	Conveyor Belt Mk.2	120
Mk.3	Conveyor Belt Mk.3	270
Mk.4	Conveyor Belt Mk.4	480
Mk.5	Conveyor Belt Mk.5	780
Mk.6	Conveyor Belt Mk.6	1200

How an N → M balancer is built

The general recipe has two stages. First, merge the N inputs down onto as few trunk belts as the target belt tier allows – ideally just one. Second, build a splitter tree on each trunk that fans out to the outputs it owns, halving or thirding the rate at every level since a splitter only ever divides by 2 or 3. A clean 3 → 6 split needs one trunk (3 inputs merge fine on a single Mk.6 belt at 810/min) and a two-level tree of splitters below it:

Why 5 and 7 outputs need a loopback

Because a splitter only divides by 2 or 3, a splitter tree alone can only produce a number of equal leaves that is a product of 2s and 3s: 2, 3, 4, 6, 8, 9, 12, 16… Any other output count – 5, 7, 10, 11, 13… – doesn't fit a clean tree. The fix is to build the tree for the next such number above the target (5 rounds up to 6, 7 rounds up to 8) and loop the surplus leaves back into the merger that feeds the tree. The loop raises the trunk's rate above the net throughput – it's now carrying the real output plus its own leftovers – so a loopback design needs a bit more belt headroom than a clean split of the same size:

Uneven splits

Not every split is an even one. A line that feeds two smelters and one constructor wants 100, 200 and 300 items/min out of a 600/min supply, not three equal thirds. Switch the calculator's Rates control from Even to Manual and type the rate of every single input and output; it works out the network that hits all of them exactly.

The trick is that a splitter only ever divides a belt into equal parts, so an uneven split has to be assembled from equal quanta. 100 : 200 : 300 reduces to 1 : 2 : 3, six shares in total – so the tool splits the 600/min in two (one half is already the 300/min output), splits the other half in three, and merges two of those 100/min leaves into the 200/min output:

Two things to watch for:

• Leftovers go through a smart splitter. If the outputs ask for less than the inputs deliver, each trunk gets a smart splitter whose overflow output runs to a sink instead of backing the line up – so every output still carries exactly the rate you typed. Decide where that overflow really goes (an AWESOME Sink, storage, a second consumer) before you build.
• Direction doesn't apply. A manual plan always builds the split shape: mergers take belts of any rate, but a splitter's outputs are always equal, so the collect mirror – which turns the input-side mergers into splitters – cannot produce uneven belts. The calculator says so under the diagram.
• Awkward ratios get expensive fast. 33 : 67 doesn't reduce, so it would need a 100-way split; the calculator refuses anything above 64 shares and, where one exists, names the closest buildable ratio within 1 % of what you asked for. Rounding the rates to a simpler ratio (or splitting the plan into two balancers) is almost always the better build.

In the calculator you can click any In or Out box in the diagram itself to type that belt's rate; where two belts merely cross, one of them hops over the other with a small arc, and every belt meets a splitter or merger at one of its three stacked ports, so a fan-out reads as a fan-out.

Belt tiers are a hard limit

1200 items/min on a Mk.6 belt is the fastest a single belt can ever move, splitter tree or not. Once the total rate divided across the inputs would push any single belt above that – many inputs at a high rate, or very few outputs carrying a lot each – no amount of clever splitting fixes it on one trunk; the plan needs multiple trunk belts instead, and the merge stage has to grow to match. The balancer calculator works this out automatically for any input/output count and rate and will refuse a plan (with a suggested fix) instead of drawing something that can't be built.

In the planner

The planner builds balancers too, in two places. A dedicated Load balancer node (Logistics category) takes N inputs and M outputs (1…16 each) and solves as an ideal balancer in the factory graph itself – place it, wire belts to its ports, and edit its N/M/rate right there with the same tool as this page, embedded in the node's dialog. And every machine, station and generator node has a Balance belts tab next to Settings: one row per ingredient or product, pre-filled from the graph's last solve, so you can check the splitter/merger plan for that item's belts without leaving the node you're already editing. On a public world you're viewing read-only, the node and the tab are both visible but not editable.

Common layouts

Fifteen frequently needed balancers, split mode, with the belt tier picked automatically for one saturated Mk.3 belt per input. Open any of them in the calculator to change the rate, tier or direction.