Each machine has its own characteristics for specific application requirements. Which offers the best match between needed output and necessary investment?
For electronics manufacturers deciding to invest in new SMD placement machines, cost of production (CoP), operational flexibility and technology capabilities are important drivers to take into account. If we look at available SMD placement machine concepts, all machines have typical characteristics to fulfill these main drivers to a greater or a lesser extent.
A method for characterization of SMD placement machine concepts is discussed here to help manufacturers find the right machine for their production environment. An SMD placer main architecture is illustrated in Figure 1. To characterize all available SMD placers, the movement of PCBs and/or components is used as an identifier (Table 1).
Split axis concept X-1 This machine concept consists of a PCB that can move in x direction and (usually) 2 overhead gantries that move in y direction. These gantries are equipped with a number of placement heads. In most cases (vision-in-thefly) the component alignment (CA) module is installed between the feeder pick locations and placement locations. Feeder banks/feeder trolleys are located on both sides of the gantries. In this concept gantry y1 is picking, while gantry y2 is placing, and vice versa.
Advantages of these systems include an absence of limitations in the number of feeders to be accommodated, as well as picking and placing of components simultaneously. Both PCB servo system and gantry servo system are fitted directly to the base of the machine. This ensures the lowest moving mass and better accuracy.
Disadvantages include relatively long gantry strokes which have a negative influence on output, movement of PCB during placement leading to acceleration forces on premounted components, and the requirement for intelligent motorized feeders for pick corrections for small size components. The matrix tray feeder requires component shuttle or line-by-line indexing. The machine occupies a large footprint in a flowline arrangement due to its depth.
Split axis concepts Y-1 and Y-2 Concept Y-1 consists of a PCB is moving in y direction and 1 or 2 overhead gantries that move in x direction. In this concept, gantry x1 is picking, while gantry x2 is placing, and vice versa. There are 2 possible ways to integrate this machine concept into a flow line: in-line or on-line “Tmode”.
In concept Y-2, 2 PCBs are inside the placement machine. The machine consists in fact of 2 machines “under one roof”. One is located at the front side (slides: X1 and Y1), while the other is located at the rear side (slides: X2 and Y2). If the PCB at the front side is finished it should be transferred to the rear side. This PCB transport is synchronized with the PCB at the rear side (this PCB will leave the machine).
While both concepts offer similar advantages and disadvantages as X-1, concept Y-2 offers the possibility of high output per square meter. However, PCB handling is complex and time consuming. Feeders are required on both sides due to its twosided operation. In the case Y-1, when two shuttles are used on a common guiding rail, collision avoidance is required and this leads to lower output. Also, PCB run-in/run-out is time consuming.
Turret concept The turret concept is the most applied concept for chip shooting, with an output under ideal conditions ranging from 40 to 53kcph. The machine consists of an indexing wheel (turret) with placement heads (typical: 12 to 24 heads) at the circumference. In most cases each head is equipped with 3 to 6 nozzles that can be changed onthe-fly.
The feeder table is located at the rear side of the turret. It can move along the machine in x-direction and is driven by a servo-system. As a result the required feeder is positioned such that its position corresponds with the pick position of the placement head on the rear side of the turret. At the same time, the opposite placement head will place a component on the PCB (at the front side of the turret). The PCB is located on an x, y table that takes care of positioning the PCB to the needed placement positions.
The positions between the pickposition and the place-position on the turret are used to perform auxiliary operations like SMD presence check, pre-rotation, visual inspection, component alignment and final rotation. The benefit of this concept is that all these actions can take place simultaneously. In some cases the feeder table can be split into two parts (split feeder bank concept). In this case one table is moving and is used to offer components to the placement heads, while the other feeder table is prepared for fast change-over to the next batch.
However, there is a large difference between optimum output and real output with turret machines. With output determined by both turret index movement and PCB x, y movements, the latter is the determining factor in most cases. PCB exchange time is also relatively long at between 3 to 5 seconds. Subsequently, placement of large components further limits output.
Other disadvantages include the long machine length, due to the moving feeder table. This makes feeder replenishment during operation impossible as well. The vibrations caused by the feeder table movement and PCB movement imposing acceleration forces on pre-mounted components limits accuracy. Tray feeding is also not possible, and pick corrections for small size components are only possible with intelligent motorized feeders.
Single-gantry concepts 1 and 2 These are the simplest low cost placement machines with relatively long PCB exchange times of 2 to 3 seconds, and are not suited for short line cycle times. The machine consists of one or two gantry type robots, equipped with (in most cases) multiple placement heads. A second singlegantry robot added can place components on a second PCB within the machine. In this case, a relatively high output per square meter is possible.
The robots can move in the x, y surface, while the placement heads can make q,z movements. In most cases, the placement robot is driven on one side of the x-beam by means of a ball lead-screw (T-drive). A disadvantage of this T-drive is the so-called “wag tail” effect on the opposite side of the x-beam. Therefore, for high precision applications the placement robot is driven on both sides of the x-beam by means of linear servomotors (H-drive).
Component feeders are located on both sides on banks or trolleys. Feedertrolleys are growing in popularity because they allow for a quick feeder changeover. In most cases, CA modules are located on both sides of the machine between the pick-andplace locations. The PCB transport is located in the center of the machine, with the PCB either clamped on the edges or located by means of locating pins.
These concepts allow feeder replenishment during operation. Its flexibility makes nozzle exchange onthe- fly, and the accommodation of different cameras and multiple pipettes possible. In the case of multiple heads at the same pitch of feeders, gang pick increases output. Pick corrections can be done by robot, with no intelligent or motorized feeders required. The machine accommodates a large number of feeders, with tray feeders and exotic feeders easily integrated. As there are no acting acceleration forces, PCB and feeders are stationary, improving accuracy.
Disadvantages include an output limited to approximately 10kcph. Also, feeders located on both sides of the machine require a two-sided operation. Adding a second single-gantry robot reduces the number of available component codes by a factor of 2 in the case of double placement, and PCB transport takes more time because of two-fold PCB handling. Also, acceleration forces act on pre-mounted components on PCB during placement corrections, and accuracy is limited by the complexity of the corrections.
Dual-gantry concepts 1 and 2 In this concept, two gantry robots are sharing the same workspace above the PCB that is in the work area. While one robot is picking, the other robot can be placing. This almost multiplies the achievable output by a factor of 2.
Reasons for the theoretical output being always less than twice the output of the single-gantry robot is due to the synchronization of robot movements to avoid collisions, and also the balancing of workload between both robots.
Advantages of this concept include the reduction of placement cost due to sharing of base, transport and controller by both robots. H-drive can be implemented for high placement accuracy. However, this accuracy is worse than that of single gantry because of vibrations induced by the second robot. Otherwise, advantages and disadvantages are similar to that of single-gantry concept. For dual-gantry concept 2, the PCB transport system is divided into 2 parts, each capable of transporting the PCB in x and y directions. Once a PCB enters the machine it will be clamped and shifted towards the front or rear robot placement area, thus reducing the distance between pick-and-place locations per robot to a minimum.
After completion of both placement cycles, both PCBs will be shifted towards the center of the machine, where both split transport systems will be aligned to enable transfer of the right hand PCB to the left hand PCB transport system and (at the same time) move the left hand PCB out of the machine. At the same time a “fresh” PCB will enter the machine at the right side. As both robots have their own operating space, collisions will be impossible. At the same time both PCBs are stationary, which is beneficial for high placement accuracy.
While this system can yield an output as high as 40kcph, the tradeoff is the time it takes to transfer the PCB between both work areas. Collect-and-place concept 1 and 2 Each machine consists of two pickand- place robots, equipped with a multi pipette revolver head. While robot A is picking up components on all available pipette positions sequentially, robot B is placing them sequentially. The pick cycle needs to be balanced with the place cycle to obtain optimal output. Due to this principle, the relatively long travel times between pick-up locations and placement locations are divided by the available number of pipette positions. A multiple pipette beam on a single- or dualgantry concept placer can also achieve this effect of “time sharing.” In fact most dual-gantry concept placers with multiple placement heads on the gantry can also be considered as a collect-and-place category machine.
In concept 2, each machine consists of two times of two pick-and-place robots. Both opposite robots are seen as a separate SMD placer. Each machine part has a PCB in its working area (total of two PCBs in each machine).
These systems share many of the same advantages as the dual-gantry concept. As for disadvantages, there is again limited placement accuracy due to vibrations and T-drive. Revolver heads with radial pointing pipettes are more vulnerable due to space limitations, and nozzle exchange is only possible “one-by-one.” The index movement of the revolver head induces extra acceleration forces on the nozzle tips, reducing the attainable output while limited vertical stroke of placement heads on revolver lead to constraints on component height (less than 6mm). Real output is much less than theoretical output due to “nozzle starvation” as a result of com-ponent/nozzle relationship restrictions. In addition, the depth of the machines give rise to accessibility problems.
Multiple-pick & Multiple-place concepts A number of PCBs are lined up in a transport row and are indexed simultaneously over the same index stroke. A number of placement robots, independent from the number of PCBs in the machine, are located over the length of the machine. Each robot is equipped with a single nozzle placement head and is able to pick components from a limited number of feeders located at the front side of the machine. After picking, each robot will move the component to the correct placement location above the PCB. During travel, the component will be aligned (“visionon- the-beam”) by a laser align sensor. In this case, component alignment will not add time to the pick-and-place cycle. As a result the above-described concept makes it possible to pick and place components in parallel by application of a number of parallel operating placement robots. Because the PCBs are indexed, every area on every PCB will be covered by one of the placement robots at every moment. The total PCB area is thus covered in a number of index steps.
Advantages of this modular machine concept include both output flexibility and component-type flexibility. Pickand- place actions are performed in parallel, leading to possible outputs of up to 120kcph within a small footprint.
The machine is well suited for short line cycle times as feeder replenishment is possible during operation and PCB transport does not add time in the pick and place cycle. Because no acceleration forces are acting on the PCB, high accuracy can be maintained. Investment can be incremental in line with production volume increase while CoP remains stable as a function of the number of placement robots.
The system also exhibits several disadvantages. The number of feeders relates to number of accommodated placement robots on the machine, and is not independent of output. The output can be sensitive for component clusters on PCBs. Introduction of new PCB types is more complex due to the fact that there are multiple PCBs in the machine.
Machine calibration is also more complex than in machines with only 1 PCB, although applying intelligent calibration techniques can reduce this effect.
SMD placement machine concept analysis In order to compare the discussed machine concepts, a generic timing model was built using the same parameters per machine. With this timing model, it is possible to compare the outputs from the different machine concepts and to study the influence of system parameter variations on the output. The system parameter assumptions in this model are as follows:
• Servo movements on pipette: Third order speed profiles with Amax = 20m/s2, Vmax = 1.5m/s and Adot = 1000m/s3 • Servo movements on PCB: Third order speed profiles with Amax = 7m/s2, Vmax = 0.75m/s and Adot = 1000m/s3 • Pick time: 60ms per component • Place time: 60ms per component • PCB width: 500mm • Feeder width: 16mm • Number of feeders per machine: 100 • PCB transport time: 2s • PCB alignment time: 1s • Number of components to be placed: 200 • Parameters to be varied: Number of placement heads (pipettes), distance between 2 adjacent placements, Amax, Vmax
Results from the timing model Split axis concept X-1 From Figure 2 and Figure 3 it can be concluded that the number of placement heads per shuttle has the most significant influence on the output. The influence of Amax and Vmax is less significant. The reasons for this are that a high Vmax is only helpful for the long strokes between the pick area and the place area. During picking and placing of components this Vmax will not be reached. A high value for Amax only contributes partially, since allowable values for Amax on the PCB are much less due to possible shifting of pre-mounted components in glue/solder paste on the PCB.
The effect of “flattening” of the output in the case of more than 8 heads is a result of the reduction of the “time sharing” effect. Here the influences of individual pick-andplace times become dominant.
Split axis concept Y-1 Figure 4 reveals that the output of concept Y-1 is comparable with the output of concept X-1 (figure 2). In case of T-mode arrangement, the output of concept Y-1 will be less than the output of concept X-1 because of longer PCB run-in/run-out times.
Turret concept The theoretically attainable output of a turret is limited by the maximum allowable radial acceleration at the nozzle. The components on the nozzle will experience an acceleration force as a vector resulting from both centrifugal force (= m*w2*R) and tangential force (= m*wdot*R). This resulting force vector acting on the component can only be kept in equilibrium by means of a frictional force between component and nozzle.
This friction force is limited by the suction surface of the nozzle and the friction coefficient between component and nozzle. In practice, the maximum allowable acceleration between component and nozzle is limited to 50m/s2 nominal value. Therefore, the diameter of a turret needs to become smaller when a higher output is required.
Another limiting factor for the output is the maximum allowable acceleration force on the pre-mounted components. Here a maximum value between 7 and 10 m/s2 (nominal) is realistic. There will be 3 time-determining cycles in the timing diagram of the turret: 1. Turret index movement acceleration limitations 2. PCB movement acceleration limitations 3. Feeder index movement limitations (depending on displacement between 2 feeders needed for 2 successive picks)
The longest time of these three cycles will be the limiting factor for the output of the turret. This is illustrated in Figure 5, which shows that a distance between 2 placement positions on the PCB of 3cm will cause an output reduction of about 50 percent. Also, large (heavy) components will reduce the turret output. This is the reason that large components will always be placed at the end of the placement cycle, after all smaller components have been placed.
Single-gantry concept From Figure 6 and Figure 7 it can be inferred that the number of placement heads per gantry has the most significant influence on the output. The influence of Amax and Vmax is less important as a high Vmax is only helpful for long strokes between pick area and place area. During picking and placing of components, Vmax will not be reached. The effect of “flattening” of output when more than 8 heads are present is a result of reduction of the “time sharing” effect; here the influences of individual pick-and place times become dominant.
Dual-gantry concept From Figure 8, it can be concluded that the output of a dual gantry is approximately twice the output of that of a single gantry. In reality it will be slightly less due to collision avoidance between the two gantries.
A typical timing diagram of a dual gantry is shown in Figure 9. Abalance between the movement of both gantries is needed to avoid collision (Twait ), causing a slight output reduction.
Collect-and-place concepts In Figure 10, a “flattening” in output can again be observed as a function of an increasing number of pipettes on the collector wheel. In reality a maximum of 12 pipettes is used.
Multiple pick-and-place concept The timing model of the multiple pick-and-place concept can be compared with that of the single-gantry concept with only one head per gantry. Since the picking range is limited to 6 to 10 feeders per robot, the travel distances between pick-and-place area will also be limited. This has a positive influence on the output. Cycle times needed for PCB runin/ run-out and PCB alignment do not contribute to the total cycle time, since these actions can be done in parallel with pick-and-place. Therefore, an output of 5000 to 7500 components per hour per robot is realistic.
Total machine output is the output per module multiplied by the number of robots installed per machine. This modular machine set-up enables the user to invest in small steps, such that the number of placement robots installed in the machine matches the required production capacity. This has a positive influence on the CoP as illustrated in Figure 11.
The reasons for the lower CoP of the multiple pick and place concept (MPP), compared to the turret concept and the collect-and-place (C&P) concept is that the investment for the MPP concept consists of two parts—a fixed part consisting of machine base, PCB transport, machine controller and board vision module, and a variable part consisting of placement robots, feeders and nozzles.
With the MPP concept it is possible to increase the machine output and investment in small increments, leading to a “flat” CoP curve as illustrated in Figure 11. Other machine concepts, like the turret and the collect-and-place concept, can increase the machine output in large steps only (whole machine increments!), leading to higher CoP (see “saw-tooth” curves in Figure 11).
It can also be shown that the single-gantry and dualgantry concepts are preferred when a lower output is required (approx. < 35000 cph), while turret, collectand- place and MPP concepts show better CoP performance above this cross-over point.
Conclusions Machine concepts where the PCB is not moving during placements are preferred because of: • Potentially better placement accuracy (no accelerations acting on pre-mounted components). It is expected that placement accuracy will become a more decisive factor in the near future due to increasing miniaturization in electronic products. • Higher output attainable. There are no limitations due to possible shift of pre-mounted components on the PCB. Machine concepts matching this requirement are single-gantry, dual-gantry, collect-and-place, and multiple pick-and-place concepts.
Machine concepts with a modular structure offer best match between needed output and necessary investments. Low-cost machines with limited output, e.g. single gantry/dual gantry concepts, offer this modularity. However, this can lead to a large number of placement machines in line, requiring large floorspace and more operators. Therefore the MPP concept delivers lowest CoP for outputs of more than 35000 cph. Modular machine concepts also offer the highest machine availability.
Machines with single sided feeding and operation at the same side are preferred because they require less line floor space and operators. To optimize output, attention must be paid to the following concepts: • Minimum travel distances. • Auxiliary actions like PCB run-in/run-out and vision processing are performed in parallel with placement actions, thus requiring no additional cycle time. • Nozzle flexibility. • Fast changeover concepts like using feeder trolleys and the elimination of PCB specific tooling.
Overall, modular machine concepts offer best upgradability opportunities for the future.
Component miniaturization is driving placement technology
Sjef van Gastel, Manager, Advanced Development, Assembléon Netherlands B.V., discusses future trends in pick-and-place technology.
EM Asia: Which key areas of technology in pick & place equipment will we see advancements? van Gastel: There are four key areas where we will see advancements: Robot technology, Placement heads, Substrate transport modules, and Vision systems.
In robot technology, there will be an increased use of linear motor technology and so-called H-drive robot concepts. Drivers for this are miniaturization of both components and related reduction of line interspacing, leading to a need for increased placement accuracy, combined with improved robot performance. Other areas include the use of lightweight robots for faster settling and better accuracy.
Placement heads cover the use of linear motors for both Z and theta movements. The drivers are better performance and placement force control, including impact force control. Especially the placement of ultra-small components like ‘01005’ chips, which require “gentle” placement in combination with high speeds. In general, 90 to 95 percent of all components are chips! Other areas include the increased use of optical sensors for the detection of components, which offer better performance compared to vacuum detection.
For substrate transport modules, there will be an increased use of flexible substrate support and board clamping. Fast changeover concepts include the increasing use of artwork recognition as a technology to determine substrate positioning. Fast substrate transport of less than one second can be achieved using walking beam transport with parallel substrate load/unload or dual lane concept.
Modular transport systems enable rapid changeover for different substrate materials e.g. organic, ceramic and flex foil. In vision systems, component alignment requires the increased use of vision-on-the-beam for elimination of image acquisition time in the pickand- place cycle. For board alignment, artwork recognition, as we have just pointed out, will be increasingly used.
In the future, look-while-place technology will allow real-time position correction and inspection during placement. Other areas of advancement include the replacement of CCD-based image sensors by CMOS-based image sensors for faster image read-out and better dynamic range.
EM Asia: How important is it to automate placement inspection? van Gastel: To avoid costly repair actions after soldering, it is essential to determine placement quality as soon as possible after component placement. This is so that wrongly placed components can be removed easily without damaging the components. This inspection can be done by means of vision technology.
Nowadays most inspection equipment for AOI consists of separate robots containing an inspection camera. This AOI equipment is placed in front of (solder paste inspection) or after (post placement inspection) the component placement machine. It is also possible to integrate these inspection cameras into the placement machine. In the case of post-solder inspection, X-ray based inspection systems that show the quality of solder joints are preferable.
EM Asia: Where do you see emerging applications for placement machines? van Gastel: Emerging markets for placement machines include mobile phones, RF modules, camera modules, RFID devices, optoelectronics and mobile display modules. For these high-end products, requirements are dictated by miniaturization. Here accuracy (up to 30 microns at Cpk 1.33) and component range (‘01005’size chips, 3D assembly of odd components & modules, placement force control, support of adhesive interconnection processes) require advanced placement equipment.
