Line Balance and Buffer Size on an SMT Line
Line balance is the gap between the rate a line can produce and the rate it actually produces once the stations have to wait for one another. On an SMT line the constraints are rarely the placement machines, which run at a published cycle time, but the printing, inspection and reflow stages around them. Buffers between stations are the practical way to absorb the short interruptions that would otherwise stop everything downstream. Sized well, a buffer hides normal variation; sized badly, it hides a bottleneck and accumulates work in process that nobody can explain at the end of the shift.
Finding the Real Constraint
The constraint is the station with the longest average cycle time, not the station that appears busiest. Printing a fine-pitch panel with a required paste inspection step can take longer than placing the parts, especially when a reprint occurs. Reflow is a fixed-time process: the belt speed and the tunnel length determine the interval between panels, and that interval is a hard ceiling on the line unless the oven is changed. Measuring the average and the spread at each station over a full shift gives a picture that the nameplate ratings do not.
Short stops matter as much as the average. A printer that pauses for a wipe every ten panels, an AOI that stops on a false call, and a mounter that runs out of a reel are all brief events, but their frequency decides how often the line starves. The right metric is the proportion of time the downstream station has no work, and it is usually visible directly on the machine’s own counters. Where that number is high, the problem is upstream; where the upstream station is blocked, the problem is downstream.
What a Buffer Actually Does
A buffer decouples two stations so that a stop at one does not immediately stop the other. Its value is measured in time, not in panels: a two-panel buffer in front of the oven covers about two oven intervals, which may be 60 seconds or five minutes depending on the product. The buffer should therefore be sized in minutes of cover at the current takt, and re-evaluated when the product or the belt speed changes. A buffer that was adequate for one product can be nearly useless for another.
Buffers do not create capacity. They convert a hard stop into a soft delay, and they allow a short interruption to be absorbed without a line stop. Once the interruption is longer than the cover, the effect is exactly the same as having no buffer at all, only with more work in process sitting on the conveyor. That is why the buffer size should follow from the distribution of stop durations, and why adding conveyor length is a poor substitute for reducing the frequency of stops.
Where to Place Buffers
The first useful place is after printing and before placement, because a printer stop caused by a wipe, a paste replenishment or a stencil change is frequent and predictable. A short buffer there keeps the mounters running through the event. The second is before reflow, so that the oven is never starved and never has to be stopped, since restarting an oven costs far more than a pause elsewhere. A third, smaller buffer after reflow protects the downstream assembly and test operations.
Placing a buffer in front of a station that is itself the constraint buys nothing, because the constraint is never starved for long. The same is true of a buffer after the constraint: it protects downstream stations but does nothing for throughput. Mapping the buffers against the measured constraint avoids the common mistake of fitting a conveyor section wherever there is floor space, which adds work in process without adding output. The same equipment record that supports preventive maintenance should carry the buffer settings, so they are reviewed on the same cycle.
Sizing Rules That Survive Contact With Production
A workable starting point is to cover the 90th percentile of stop duration at the station being protected, expressed in panels at the current takt. If a printer’s stops last up to four minutes and the takt is thirty seconds, an eight-panel buffer gives cover for nearly all of them. The number should be reviewed whenever the fleet mix changes, because a mix with more fine-pitch product will lengthen the average stop and reduce the effective cover of the same physical buffer.
Work in process is the cost of that cover. Every panel in a buffer is inventory that has to be tracked, that can be damaged during the wait, and that hides a defect until a later station finds it. Where a defect does occur, a long buffer means the fault is repeated on every panel already in the queue. This is the argument for keeping buffers short and for locating inspection immediately after the process that most often fails, rather than at the end of the line.
Balance Across Shifts and Products
Balance is not a fixed property of the line. It changes with the product, with the operator, and with the time of day. The first hour of a shift usually runs slower than the middle, and the changeover from one product to the next is a period in which every station is out of balance by design. A line that is balanced on average will still show starvation in the first hour unless the start-up sequence is planned, for example by staging printed panels before the mounters are released.
Recording the counters at the shift handover turns this into data rather than opinion. The incoming team can see whether the line ran at the expected rate and how much of the loss was attributable to each station. Over a few weeks the pattern shows whether the constraint is stable or whether it moves with the product mix, which decides whether the answer is a capital purchase or a change in scheduling.
Improving Balance Without Capital
The cheapest improvements are usually in the changeover and the setup. Reducing the number of stencil changes by grouping like products, pre-staging the feeders for the next product while the current one runs, and standardising the paste replenishment interval all reduce the frequency of stops without touching the equipment. Each of these is a change to the schedule or to the work instruction, and each shows up directly as reduced starvation at the next station.
Where the constraint is genuinely the printer, moving inspection off the critical path or splitting a print operation across two machines can release capacity that no amount of buffering will create. The decision should follow the measurement, and the measurement should include the first pass yield at each stage, because a station with a poor first pass yield consumes capacity through rework even when its cycle time looks acceptable. Tracking that yield per station, not per line, is what keeps a manufacturing yield review honest.
FAQ
How much work in process is too much? When the queue in front of a station is longer than the cover needed for its own stop distribution. Beyond that point the extra panels only lengthen the feedback loop between a fault and its discovery. A useful discipline is to cap the buffer physically, so the limit cannot drift upward without a decision.
Does a faster mounter improve throughput? Only if the mounter is the constraint. If printing is the constraint, a faster mounter simply waits more often, and the money would have been better spent on a second printer or on reducing print stops. Measure before buying, and measure the constraint rather than the average.
Should buffers be automatic or manual? Automatic conveyors are preferred where the panels are heavy or where the buffer must be traversed without operator attention. Manual racks are cheaper and allow re-inspection, but they introduce a handling step and a place for panels to be mixed. The choice usually follows the volume rather than the technology.
