This page is out of date and does not represent current practice.

Stringing Machine Purpose and Goals

The stringing machine's task is to inject loops of fiber into NOνA cells. It must do this quickly and accurately without damaging the fiber. It should be easy to use and should be able to recover well from errors.

Major Constraints

Fiber is fragile

Fiber is prone to being scratched, so we must make sure it does not move over any rough surfaces. Ideally it should never slide on any surface.

Fiber is prone to being kinked. We will be using Kuraray s-type fiber with a diameter of either 0.6 or 0.7 mm. Fiber with a diameter of d mm is rated down to a bend diameter of 100d mm. Because of the size of the NOνA cell, we will be exceeding this diameter somewhat at the end of each loop for 0.8mm or 0.7mm fiber, however we want to avoid doing it anywhere else.

We've been told that 0.8mm fiber is rated to a tension of 10N, but it is not well understood what this means. Tests done by Dan Cronin-Hennessy have shown that 0.8mm fibers can support roughly 25N before breaking, but it almost certain that damage occurs before this point. Tension tolerance presumably scales with the square of the fiber diameter. We would like to keep the tension on the fiber well below 10N at all times.

Fiber remembers its bend radius

The fiber wants to be the shape it was on the spool it was shipped on. When it is wound onto smaller spools, it must be held at a non-zero tension because otherwise it will expand and fall off.

Components and Definitions

• Main spool: the spool that the fiber comes on. It has a ~3m circumference.
• Grooved spool: the metal spool that stores half a cell's worth of fiber in a helical groove. It has a 1m circumference. It has a clip which holds the end of the fiber in place.
• Buffer: The buffers exist to store fiber such that if the tension starts to increase or decrease away from the desired value, the change is automatically mitigated by decreasing or increasing the amount of fiber in the buffer. (Otherwise the fiber would either break or fall on the floor, respectively.) Each consists of a wheel mounted to a track. The wheel is pulled upwards by a spring and is attached to a spring potentiometer to measure its position.
• Main buffer: the buffer between the main spool and the cutter.
• Grooved buffer: the buffer between the grooved spool and the puck.
• Main (grooved) motor: the servo motor which spins the main (grooved) spool.
• Top fiber: the fiber which is spooled into the cell from the grooved spool.
• Bottom fiber: the fiber which is spooled into the cell from the main spool.
• Cutter encoder: an encoder on the wheel immediately before the cutter which senses how much fiber has passed this point.
• Puck (sometimes called the "parachute" or the "ring", depending on implementation): The device which the fiber is attached to as it is sucked down the cell. So that we can keep track of the lengths of both ends of the fiber, the fiber must not slide on the puck.
• PLC: Programmable Logic Controller, a specialized computer which controls the system. It is programmed in assembly and can perform arbitrary calculations. Has digital and analog inputs and outputs.
• Servo drive: a box which controls the servo motors. It stores many parameters such as preset speeds and acceleration profiles. It takes analog and digital inputs and has digital outputs. These are used to activate presets, control velocity, handle errors, etc.
• Motor modes:
  • Tension mode: The PLC can read the position of the buffers and control the speed of the corresponding spool to keep the tension at the desired value. When the PLC is doing this, we say that the spool is in tension mode.
  • Position mode: The servo can direct a spool to turn a specified number of times and then stop. In this mode, that spool does not respond in any way to the tension on the fiber. (Since there are two spools and buffers, one can be in position mode as long as the other is in tension mode.)
  • Velocity mode: The servo can direct a spool to turn at a constant rate. This is used in two ways:
    • Jog: During setup, the operator can jog either spool forwards or backwards at a fixed low speed. This is primarily useful for getting the fiber into the main buffer (when a new spool is being started or an error has occured).
    • Grooved spool clip alignment: The grooved spool needs to start each cell with its clip at the top. At the end of spool out, it will release the fiber when the clip is in this position, but the spool will not stop instantaneously at this time. So after each cell it will turn until it senses that the clip is in the right place. This is also useful for error recovery. Whenever the system resets, it will align the clip.
  • Stopped: When the PLC is neither controlling the speed nor position of a spool, but the servo is enabled, the motors will not turn. (The motors also have brakes, but they aren't necessary in this case.)
• Touchscreen: The touchscreen controls the PLC by changing values in its memory in response to button pushes. The touchscreen cannot perform arbitrary commands when buttons are pushed, not even to the extent of changing two PLC memory locations in response to a single button push. All such functions must be done in the PLC. However, it can choose what buttons to display based on the PLC state. This is used to avoid allowing the operator to do things in the wrong order.

Stringing Machine Operations

This information is summarized in a flow chart.

Order of operations in error-free stringing

C1: Ready to start. System is finished with the previous cell and is ready to start another. The main spool is in tension mode. The grooved spool is stopped. The grooved buffer is locked.

E1: Operator takes hold of the fiber, releases it from the cutter and readies the cutter to cut again.

C2: Threading. Operator pulls the fiber through the machine.

E2: Operator attaches fiber to the grooved spool.

C3: Waiting to spool up. Fiber is attached to grooved spool.

E3: Operator pushes SPOOL UP. The system engages position mode on the grooved spool.

C4: Spooling up. The system pulls the right amount of fiber onto the grooved spool for the current cell.

E4: The grooved spool stops.

C5: Spooled up. The system is ready for the fiber to be attached to the puck.

E5: The operator attaches the fiber to the puck. The tension on the two sides of the puck is now independent (ideally).

C6: Fiber attached to puck. The system is ready for the grooved buffer to be unlocked.

E6: The operator unlocks the grooved buffer.

C7: Almost ready to string. The system is ready to have STRING CELL pushed.

E7: Operator pushes STRING CELL. The system engages tension on the grooved spool.

C8: Ready to string. Tension is on, puck is stationary.

E8: Operator begins to move puck.

C9: Pre-stringing. Operator pulls puck towards cell.

E9: Operator inserts puck in cell.

C10: Stringing. Puck and fiber are sucked down the cell.

E10: At roughly the same time, both ends of the fiber are released from the machine. The top fiber is released from the grooved spool when there is no more fiber there. This is accomplished mechanically. The bottom fiber is cut when the cutter encoder senses that the right amount of fiber has passed. The end of the bottom fiber which is still connected to the spool is clamped in the cutter as part of the cutting motion.

C11: Post-stringing. Puck and cut fiber continue freely down the cell for a short time. The vacuum turns off a few tenths of seconds into this. Main spool handles maintaining tension on the fiber which is on its side of the cutter. Grooved spool re-algins its clip to the top.

E11: Operator locks grooved buffer and moves machine to next cell. Machine makes SPOOL UP button available.

State table for error-free stringing

Italics indicate a change in state.

Time estimates

Based on a video taken on 2 Jan 2007 in which I (Matthew Strait) strung three cells:

• E1-E3 (threading): 11s
• C4-E4 (spooling up): 12s
• E5-E6 (puck and buffer handling): 6s
• E8-E11 (stringing and setup): 18s

Total per cell: 46s. This does not include the time it takes to attach the fiber to the puck, since no mechanism for that was available on 2 Jan 2007. Adding a few seconds for this and rounding up for safety gives an estimate of 60s. Realistically, it would not be surprising if a 45s average were achieved by machine tweaks and experienced workers, but at this point no promises can be made. Adding 5 minutes overhead per module since a new spool will be needed roughly once per module gives 37 minutes per module total for stringing.

Problems and Solutions at Specific Steps

C1-C2: Ready to start, Threading

• If a new spool is being started, the main buffer will have no fiber in it. In order to get started, the operator will use jog mode to advance the fiber at a constant speed until there's enough in the buffer that the tension can be engaged. No problem.
• The fiber might slip out of the cutter or the operator might lose hold of the fiber. The main buffer will sense a complete loss of tension and trigger the motor to stop. The system should then offer to try the same cell again. It would be nice if the operator had a fiber holding device rather than using bare hands. This would make slipping less likely and also reduce the likelihood of fiber damage through careless handling. Not critical: to be worked on.

C3-C4: Waiting to spool up, Spooling up

• The fiber can come undone from the clip before or during spool up. The main buffer will sense the loss of tension and stop both motors. Any fiber that is spooled up will likely fall on the floor and should be discarded. Doesn't seem to happen.

C4: Spooling up

• If the grooved buffer is not locked, the wrong amount of fiber would be spooled onto the grooved spool. The system should not display the option to spool up unless the grooved buffer is locked. Done: no problem.
• The fiber does not always spool onto the grooved spool cleanly. Sometimes it skips a groove, for instance. This can't be sensed automatically, but the operator should be watching. What then? How serious is this? To be worked on.

C5-C9: Spooled up, ..., Pre-stringing

• The grooved buffer must be locked when the fiber is attached to the puck. It must be unlocked when the fiber is being pulled out of the machine. Since the system cannot detect the attachment process, the operator must do the right thing. No problem.

C10: Stringing

• The vacuum must not be so strong that the tension mode can't keep up with it. If either buffer bottoms out, it would be nice if the vacuum could be automatically throttled. If that's not implemented, the system should at least display a warning. Doesn't happen much: to be worked on.

E10: Fiber release

• The clip may fail to release the fiber from the grooved spool. The fiber will likely be damaged, so this is probably not a recoverable situation. The operator should throw it away and try again. This occurrence may be automatically detectable because if the clip manages to keep hold of the fiber through the next half rotation it will quickly bottom out the grooved buffer. (The spool will be pulling the fiber backwards out of the buffer and the buffer will respond by spinning it faster in the same direction, since it expects the fiber to only ever be wound the other way!) On the other hand, if the fiber gets ripped out of the clip before this can happen, it will be up to the operator to notice the problem. Doesn't happen much, if at all.
• The cutter does not engage or engages but doesn't cut. This is recoverable. The operator does the job of the cutter and cuts off a liberal amount of fiber which can be trimmed later. No problem.

C11: Post-stringing

• The puck bounces off of the vacuum machine at the far end far enough back into the cell that the person at that end can't reach it. Probably not a big deal as long as it is infrequent, since the vacuum can be used to suck it back. Taken care of by tuning the vacuum timing.

Other Concerns

Finiteness of fiber spools

The main spool will run out of fiber roughly once per module. How will this happen and how do we handle it? Will it lose tension significantly before the end, or is it taped down? The effect of losing tension, either because we are close to running out of fiber or because we have run out is that the main buffer will lose tension. This can happen at any step of the stringing process. In general, the correct response is for the system to shut off the motors and wait for the operator to reset it. This might be a little clumsy if the cell is half-strung, so maybe some elegant way of recovering from that should be developed.

Should the system keep track of when the main spool is likely to run out and warn the operator somehow? (I think this is probably unnecessary.) How well determined are the lengths of fiber on the Kuraray spools?

Accuracy of fiber length

The length of the fiber that is spooled onto the grooved spool is precise better than 1mm. The error on the other side of the loop is much greater because it has to be cut while it is in motion. The process of attaching the fiber to the puck (by hand) also introduces substantial error.

The goal is to waste the least fiber. If it is too short, the entire 30m is wasted. (Unless we accept some fibers just not reaching quite all the way.) If it is too long, then the length over the optimum length is wasted. So if the length of the fiber is a Gaussian random variable with mean μ longer than optimum and a standard deviation σ, then the average wasted fiber for a cell is:

 Integrate[(30-x)/Sqrt(2π) e^{-(x-μ)2/(2σ2)}, {x, -∞, 0}]
+Integrate[   x  /Sqrt(2π) e^{-(x-μ)2/(2σ2)}, {x,  0, ∞}],

Assuming 500,000 cells, let's look at the effects of various degrees of precision of the cutter:

A plot of μ vs. fiber wasted shows, not too surprisingly, that it's best to err on the side of larger μ, since the function rises gently towards large μ and sharply towards small μ. Factoring in the cost of time taken to unstring and restring cells also pushes the optimal μ upwards. Most likely we will not be able to achieve sub-centimeter precision, so we should plan on wasting at least 20km of fiber due to this.