In a myriad of manufacturing operations, the initial material is stored on large spools or as long pieces of stock. Paper, plastic, metal foil, and fabric are often handled this way, fed from rolls to a machine that processes it to create end product. The continuously fed material is called the “web,” and the entire operation, “web processing.”
Perhaps the most traditional example of web processing is that of paper printing. Paper unwinds from large spools while rotating ink drums print images onto the moving target. Other common web processes include punching, stamping, and cutting.
In most cases, material flow is uninterrupted, even during printing or cutting. A continuous flow is preferable to starting and stopping because it achieves higher throughput. More products per minute mean more dollars to the manufacturer. Continuous processing also use less power and is easier on components because everything moves relatively smoothly.
In a start-stop situation, a lot of power is required to accelerate and decelerate the web; constant starting and stopping also causes machine components to wear faster. A machine operated this way is more expensive to build and operate, and does not last as long as one with continuous flow.
Processing material as it moves requires synchronization. In printing, for example, the surface of the print drum must move at exactly the same speed as the web in order to avoid smearing. Similarly for cutting, the knife edge must move at the speed of the web in order to avoid tearing or bunching.
Synchronization is achieved by “electronically gearing” the controlled motion to some measured motion. In printing, for example, an encoder may be used to measure speed. A controller then uses this information to regulate the print drum’s speed so that the print surface is in synch with the paper while they are in contact.
In most cases, web-processing machines must handle several different product lengths. In the printing example, if the length of the required print cycle matches the circumference of the drum, then the drum can rotate at a constant 1:1 ratio with respect to web travel. But if the print cycle is less than or greater than the drum circumference, then the drum must advance or fall back while it is not in contact with the paper. This print length compensation must repeat every print cycle, necessitating the use of a repetitive cam profile.
The key difference between the method discussed, called “following,” and other methods of synchronization is the assignment of a “master axis.” The motion of this axis is measured (as with an encoder) and referenced by the follower.
Ratio versus speed
The concept of ratio, the change in axis position with respect to the change in master travel, is at the heart of following. Ratio is analogous to velocity, except that the follower moves as a function of master motion instead of time. In the same way that a motion axis might have to speed up or slow down (change velocity), the electronic gearing ratio might have to increase or decrease within a profile.
A change in ratio is, of course, analogous to acceleration. As long as you know the rate of change — the total change over a given region of master travel — you can be sure of the position relationship between the master and follower axes.
The master cycle concept provides a useful way of dividing continuous master motion into meaningful portions. This allows master travel to be measured and referenced in terms of cycles and positions within a cycle. Because master cycles usually relate to product cycles, it is important to begin the measurement of master travel at a spot corresponding to the start of the product cycle. This is usually achieved by detecting the arrival of a product or moving machine part with an electronic sensor.
In most applications, one of more machine parts must maintain a 1:1 travel ratio with other moving parts. But there’s more to it than merely trying to match speed. The related parts must also be properly aligned or “in phase.” For this reason, at various points within a cam profile, the controller will advance or retard the follower. This procedure, called a “phase shift,” can put two axes in synch without affecting the ratio of the motion.
Continue on page 2
Ordinarily, data describing the shape of the cam profile is incorporated in a motion program and executed by the motion controller. This program typically contains the positions and slopes of a small sample of points along the profile. The controller generates intermediate values through interpolation based on stored programs and measured input.
Cam profiling methods
Some web processing applications involve short cycle times with several ratio changes during each cycle. In general, faster cycle times mean higher throughput.
One approach to implementing a profile with changing ratios is to write a program that monitors master position and commands ratio changes throughout the master cycle; however, it may not be possible to execute the individual commands required to change ratios and still meet cycle time requirements. The answer here is to predefine master and follower position relationships and ratios. In other words, the profile is defined before it is actually run, and it is saved in a compiled form.
In this sense, it is very much like a single complex move command. Only a single event is required to start the move, and when the profile is actually run, it is not affected by the execution speed of concurrent program commands. Such a profile is called a cam table.
General-purpose cam tables may be built from data about specific master positions and corresponding follower positions. For web processing applications, data represents web cycle positions and corresponding follower positions.
In some electronic camming functions, the controller does a simple linear position interpolation between adjacent points to create a profile. Many pairs of points are required to achieve a smooth change in ratio. Other camming functions perform quadratic or cubic interpolation, requiring fewer points for a smoother profile.
Alternatively, a profile may be constructed with sequential motion segments. After master travel has exceeded the end of the table, the profile starts again at the beginning. These profiles may be executed as one-shot moves, or as repetitive cycles, as appropriate for the application.
Progress through the profile may take place either forwards or backwards, following the master’s direction. If the profile is to be executed as a repetitive cycle, it may be possible to designate one or more “leadin” segments to precede a repetitive pattern. Once master travel has gone beyond the lead-in segment(s), repetition will take place only within the repetitive cycle, even if the master moves backward.
In many cases, there will be an existing pattern on the web, and each cut or print must be located correctly with respect to that pattern. Because webs stretch and slip, a registration mark on the web must be detected each cycle, and any correction required for perfect registration must be applied to the follower. The correction is applied in the form of a phase shift, subject to a maximum correction per cycle and a tolerance of phase error.
For example, assume registration marks occur once per master cycle at a specific position. By looking within a window around the expected registration mark, the controller can ignore all other marks printed on the web.
With each registration mark it senses, the controller corrects its own concept of the master cycle position, and advances or retards the follower by the amount required to synchronize the web and follower. If no mark occurs within a window for a given cycle, or the error is within the tolerance, then no correction should take place during that cycle.
When registration marks fall within the expected window, the controller records the actual master cycle position and compares it to the anticipated position. If the resulting error is larger than a pre-defined tolerance, the controller immediately calculates the amount of shift required to align the follower with the master cycle.
The specification of a tolerance allows the controller to ignore insignificantly small variations. If the error is larger than the maximum allowable adjustment per cycle, only some of the error is corrected. The remaining error, detected at the next registration mark, is further reduced, perhaps entirely. Limiting the amount of adjustment per cycle means that sudden errors are gradually, rather than abruptly, corrected, reducing wear and tear and possible web breaks.Continue on page 3
Factors affecting accuracy
Several accuracy requirements unique to following applications deserve special attention. For starters, the follower must maintain positioning accuracy while in motion, not just at the end of moves, because it is trying to stay synchronized with the master.
Overall positioning accuracy depends on several factors. Just as with a mechanical arrangement, accuracy errors can build with every link from beginning to end. The overall worstcase accuracy error is the sum of all errors.
Errors fall into two broad categories; master measurement errors and follower errors. Both ultimately affect follower accuracy because the commanded follower position is based on the measured master position.
Master resolution — The bestcase master measurement precision is the inverse of the number of master steps per user’s master unit. Even if all other sources of error are eliminated, follower accuracy will only be that which corresponds to one step of the master.
Follower position calculation latency — The motion controller measures master position, using it to calculate and control the corresponding follower position. The functions of measurement, calculation, and control all take time. During that time, the master is still moving, so by the time control is implemented, the calculated position is obsolete. Some controllers compensate for this latency.
Velocity smoothing — This helps reduce error due to sampling accuracy, but it increases error due to variations in master speed when latency compensation is on. Most applications maintain a constant master speed, or change very slowly, so the effect is minimal. But if the master is changing rapidly, there may be a significant master speed measurement error. This error is about one half the master speed change over the velocity smoothing sample period.
Follower resolution — The best-case follower precision is the inverse of the number of follower steps per user’s position unit. Even if all other sources of error are eliminated, follower accuracy will only be that which corresponds to one step of the follower. This must be at least as great as the required precision.
Follower tuning — Precision also depends on how accurately the drive follows its commanded position while moving. Even if master measurement were perfect, if drive tuning is poor, precision will be poor. The better the drive is tuned for smoothness and zero following error, the better the positioning precision . Often, this only matters for a specific portion of the profile, so the drive should be tuned for zero following error at that portion.
Accuracy of load mechanics — The accuracy (not repeatability) of the load mechanics must be added to the overall build up of accuracy error. This includes backlash for applications that involve motion in both directions.
Repeatability of high-speed inputs and sensors — Some applications may use high-speed inputs for functions like registration. For these applications, repeatability of highspeed inputs and sensors add to overall position error. The product of velocity and time equals distance, so the error due to repeatability may be significant at high speeds.
One final caution regarding accuracy: In many applications, master and follower units will be the same, but the number of master and follower steps per unit may differ. For example, if there are 500 master steps per inch of material, an error in master measurement of one encoder step translates to 0.002 in. of follower position error. If the ratio is not 1:1, follower error is simply master error times the ratio of the application.
The price of speed
High speed means high throughput, and this improves the bottom line. But achieving that speed comes at a price.
Several issues must be considered with respect to increased web speed. The first is the relationship of speed to position error. As speed increases, a machine may require higher quality sensors, mechanics, and more sophisticated controls in order to maintain the desired precision. All of these add cost as well as development time.
The next issue is increased demand for maximum motor speed and torque. For a given maximum ratio, the maximum follower motor speed is directly related to maximum web speed. What is not so obvious is that maximum acceleration and torque increase with the square of the increase in web speed. This is because the ratio profile is based on master travel, not time.
Continue on page 4
As web speed doubles, the change in follower speed during an acceleration ramp also doubles. But the time over which the doubled speed change takes place is cut in half. Because peak torque requirements are directly related to peak acceleration requirements, the peak torque of the motor-drive combination also goes up with the square of web speed. Motor-drive combinations with higher speed-torque ratings cost more.
Power dissipation requirements, of course, also increase, in this case, according to the fourth power of web speed. Power dissipation in the motor is proportional to the product of the winding resistance and the square of the current. And because current is proportional to torque, which goes up with the square of web speed, power dissipation increases with the fourth power of web speed.
The practical implication of power dissipation is heat dissipation. At slower web speeds, simply bolting the motor to the machine’s case may be adequate. At higher speeds, it may be necessary to add heat sinks and fans, or even water-cooling.
Following repetitive cycles
Web processing profiles generally have a section of constant ratio during which the printing or cutting takes place. The master travel and location within the master cycle of this section are important. The remainder of the profile is simply whatever it takes to get the follower back to the correct master cycle position.
For this type of profile, it makes more sense to focus only on critical points (transitions and end points) rather than the entire profile. To simplify programming, all profiles should be specified in terms of machine and product parameters as opposed to unintelligible master-follower coordinate pairs.