What was traditionally called ‘pick and place’ has two primary functions to satisfy in order to ensure a high quality placement process:
These two seemingly straightforward tasks are getting ever more demanding and complex as the diversity of components grows and as footprints continue to get smaller and more closely packed. Components such as chip capacitors and MELFs are getting smaller, while others like QFPs and TABs are getting larger, as their lead counts increase:
These trends are the reason behind increasing equipment complexity and specialisation.
By contrast with through-hole components, which we shall be dealing with later in this unit, parts for surface mounting use just a single placement method, where the part is removed from its carrier by a vacuum tool, carried to the intended location, and pressed lightly into the solder paste. Whilst many types of machines are used, they all have the same requirement, which is to have the component presented in a known position relative to its packaging. Because surface mount assembly is favoured for being a fast operation, suppliers have gone to such extreme lengths that even the most apparently unsuitable parts are now packaged in one of the standard forms.
Simplifying surface mount device shapes, it could be said that they come in three basic forms, short cylinders, rectangular blocks and flat slabs, which have to be packaged. Commonly these come from the supplier in four basic presentation formats; tube, tray, tape and cassette.
The format used most widely is tape, which is well covered by international specifications.
There is less standardisation in the field of tubes and trays, which are usually adaptations for specific devices. Some adaptations of the DIP stick (a channel holding dual-in-line packages such as ICs) have been made for small outline devices (SOICs).
Imagine yourself to be a production engineer who has been asked to buy new equipment. What would you think to be appropriate criteria for selecting feeders?
In the early period of surface mount technology, components were supplied mainly in stick magazine format and, whenever large quantities were required, in bulk format. The original bulk feeders were vibratory feeders, specially tooled to handle a given size of part. By means of a ‘stop block’ at a known point, the feeders provides a source of components for the pick-up nozzle of the placement machine.
Although still used for mechanical components on odd-form placement equipment, vibratory feeders are less suitable for surface mount components, because they:
A number of alternatives were implemented, and with most types of surface mount components the user can now choose between two or more types of packaging. Nowadays, tape format has established itself as the most important type, although fine pitch formats continue to be processed in other packaging forms.
Bulk handling of components is again gaining in importance. The factors responsible for this are the problem of waste when using tapes, the introduction of smaller and smaller passive components and the cost of the tapes themselves.
As the flexibility and placement rate of systems has increased, so have the demands made on the component feeder systems. A high product mix and correspondingly small batch sizes result in frequent feeder changing. Quick feeder changeover is required in order to minimise machine down-time, so feeders must be designed for fast replacement. One way of significantly reducing changeover times is to use feeder changeover tables, so that the process of changeover is reduced to an exchange of feeder banks.
The possibility of set-up errors will always be present, especially during high-speed changeovers. By using appropriate control systems, such set-up errors can be detected quickly and reliably before a component is placed.
‘Intelligent feeders’ are capable of storing information about the feeder type, any adjustable feeder parameters, and the component fitted, and will supply this to the placement system controller.
By this means, the machine control computer can quickly and reliably check whether the components required by the placement program are available and in the intended location. An extension to the program may also allow alternative component feeders to be accessed when a feeder runs out of components, thus maintaining production.
Tube magazines, also known as ‘sticks’, are primarily used for handling items with a small to medium pin count and without projecting leads, such as SOJ-ICs and sockets. Made of transparent, electrically conductive plastic, they are normally used for smaller volume requirements.
Tube magazine feeders feature metal guide rails which are plugged into the tubes. The components are moved to the pick-up position using a linear vibrator to assist the slope of the feeder. When a feeder is empty, there is still enough time to exchange it for a full one. With stationary feeders it is not necessary to interrupt the placement cycle.
The advantages of tube feeders compared with vibratory bulk feed are:
The disadvantages are:
The tray format of component presentation, commonly known as the ‘waffle-pack’ method, is used primarily to feed large components and those with fine lead pitches. Such trays were first seen with the introduction of QFP and DIL style devices.
Waffle trays are generally available in sizes of 25.4 mm × 25.4 mm and larger. There are also large vacuum formed packs, as large as 305 mm × 508 mm, used for flat-pack type devices. The individual internal component recesses of these trays are made to conform to the size of individual parts. As a result, a wide variety of tray sizes is available.
Goods inwards handling and storage are not a real problem, but once the trays are made ready for machine loading, in-house transportation and handling become a major concern in order to prevent component loss or damage.
Placement machines using waffle-trays have to divert their placement sequence to place other components during the time that waffle tray changeover is taking place. This can lead to an unnecessary increase in cycle times if the number of QFP component types is high. A cassette magazine can be installed which reduces the down-time for component replenishment.
Components on waffle-trays place considerable demands on the flexibility of any surface mount pick and place system since they require a lot of space and automatic handling systems if they are to process a large number of different component types and eliminate down time for tray replacement. Equipment with tray feeder towers, and able to handle large components, tends to involve a pretty substantial investment in space, as well as in capital!
The advantages of tray feeders are:
The disadvantages are:
This format has been developed from the taping of axial components, although without any need to crop leads. The most widely used format, it appear to have been based on cine-film practice with sprocket holes along one or both of the tape edges.
The first tapes were built of laminated paper, with fixing tapes top and bottom to give a three-layer construction. These have mostly been superseded by a two-part plastic tape, with a carrier tape formed to give ‘pockets’ which fit the components.
This method of presentation has now developed to include 12, 16, 24, 32, 44 and 56 mm width standards, and recent developments have introduced 200 mm wide tapes for connectors.
The most common standards for taping surface mount devices are those produced by the EIA (the Electronic Industries Association, Washington DC). The EIA-481 standard provides dimensions and tolerances for taping surface mount components so that they may be handled by all makes of placement machine.
Associated with the standards are a number of conventions:
Three materials are used in the manufacture of a tape:
Three-layer paper tapes are still occasionally used, coming mainly from the Far East. This medium is extremely difficult to work with on some European manufactured feeders. Usual difficulties are with removing components from the masking type bottom tape and with the associated handling debris. Also, because both paper cover tapes and cardboard cavity tape are hygroscopic, they absorb and release moisture depending on the ambient conditions. Plastic reels packed in plastic boxes are to be preferred.
The basic construction of a tape feeder is shown in Figure 1. Tape is unrolled from the reel into the module, where the cover strip is pulled off so that the components at the pick-up position can be lifted out by the placement head. The cover strip pulled off is wound onto a small collecting reel. The remaining empty tape runs back out of the module. Some designs incorporate an automatic cut-off of the empty tapes to relieve the burden on machine operators.
After the placement head has removed the component, the tape is indexed one increment further so that the next component can be removed at the pick-up position. Depending on the machine design, this action is either triggered mechanically by the placement head actuating a release lever on the module, or carried out electrically, using a solenoid.
In practice there are two types of indexing mechanism:
sprocket wheel –a wheel fitted with radial pins which rotates in a step-stop (indexing) movement
‘walking beam’ mechanism –a set of two pins which move with respect to each other and index the tape along
Where a placement system has stationary feeder modules, changeover and replenishing can sometimes be performed during the placement cycle, depending on access and safety considerations. Taped components can be replenished easily by simply splicing the start of a new reel onto the end of an almost empty one. The placement cycle does not have to be stopped to perform this operation. This reduces idle time and increases utilisation and placement rate.
Many components (especially chips) finish their manufacturing process in some sort of bulk packaging, and need to be transferred onto tape. Components are also frequently ‘reformatted’ for the following reasons:
The reformatting of components is thus becoming an everyday requirement. Because of the wide range of tooling and equipment required for preparing and presenting components on tapes, component suppliers are now increasingly using sub-contractors for this task, and there are now a number of specialist service companies.
The main disadvantage of the tape format is the inability to recycle the empty tapes. Especially in the case of small chip devices, the tape waste material weighs several times more than the packaged components. As a result, more components are now being processed in bulk. The components best suited to bulk feeding systems are those with rectangular or cylindrical outline, that are non-polar and do not need any differentiation between top and bottom faces. On this last count, resistors can sometimes not be used in bulk feeding systems.
Early systems were of an open format which led to frequent component loss and allowed foreign objects to get mixed with the components. These problems were eliminated by the introduction of bulk case systems, transparent containers which hold the components (Figure 2). In these, components cannot get inter-mixed or damaged and are safe from the possible ingress of any foreign bodies.
After the bulk case has been placed on the feeder, the aperture at the base is opened using the slide, to transfer parts into a cassette, from where they are transported to the pick-up position.
Because bulk components have no specific orientation, the task of the bulk component feeder is more complex than with a tape-and-reel feeder. Consequently, they are more expensive, though usually the mechanisms can be mounted on standard placement machines at the same pitch as for 8mm tape feeders.
The task of the feeder is to extract components singly from the bulk package, align them, and then transfer them to the component pick-up point. In order to achieve this, components have to be fed one at a time through a gate, and may also need to be rotated – typically rectangular components are easier to feed in the direction of their long axis, but chip pick-up usually takes place with the chip at right angles to the feeder. In consequence, some feeder designs include a means of rotating individual components. Three types of bulk feeder are available, using different methods of aligning components, using a rotating drum, a hopper, or an air blast.1
1 A short description of these methods, as well as other information about bulk packaging, may be found at http://www.murata.com/bulk/
Empty cassettes can be returned to the component supplier for replenishment. Depending on wear and condition, cassettes can be recycled five to ten times, and materials from cassettes that cannot be reused can be completely recycled.
The bulk cassette system has a number of advantages over the tape format:
Given these advantages, and the fact that a bulk case standard was set by EIA Japan in 1992, it is surprising that more assemblers are not using bulk cases. Partly this is a consequence of the lack of any international standard, but lack of pressure from component manufacturers has contributed, the case having been promoted initially by only one manufacturer (Murata), although this packaging is now widely available.
Lack of take-up probably reflects the main disadvantage of the system, its ability only to be used with a limited range of components. Another deterrent is that special component feeders are required, and these feeders are very sensitive to small changes in chip dimensions, making it difficult to mix components from different sources. However, bulk cases are with us to stay, and give particular advantages with 0402 and 0603 capacitors, where the savings can be greatly enhanced by restricting the range of components used to only the most common standard values.
Your design for a simple circuit uses thirty 0603 capacitors and resistors, and just a single SOJ-packaged integrated circuit. Which would be the best of the component packaging options available when your need is to hand-craft just a very few prototypes in-house? And how do the options change when high volume manufacture is to be contracted out?
We started our exploration of placement by considering the ways in which components are packaged. While it might seem a little bizarre, it reflects the reality that, for manufacturing engineers, the principal feature of any machine is the way in which a component is presented.
The next important attribute is the accuracy of the machine. This determines the range of components which can be correctly placed, because the positional accuracy of a surface mount device on its solder foot-print determines the shape and quality of the final solder joint. A mis-positioned component can result in bad joint reliability, in failures such as solder bridging, or even in no joint being formed at all (open circuit).
IPC-A-610 Acceptability of Electronic Assemblies reflects what is generally regarded as good practice for component placement for a very wide range of electronic assemblies. Unfortunately, these guidelines refer to the situation after soldering. This is a process which can expose some placement defects (such as those which give rise to tombstoning) yet cure others through the centring action of components ‘floating’ on molten solder.
Placement inaccuracy is the deviation of the component outline, including component leads, from the position on the circuit board of the solder footprints for that device. The maximum allowable deviation, to be given for every device/footprint, is the resultant of deviations in X, Y, and theta (q).
The placement accuracy required will vary for each different device type. Variations in the type and size of the component or the pitch of its leads will require different tolerances. There is, however, a good ‘rule of thumb’ that holds true for any SM component, which is that a component lead or terminal should not be placed more than a quarter of its width beyond the component pad, this being related to the IPC-A-610 standard.
On most fast placement machines, such as ‘chip-shooters’, the worst-case placement deviation is of the order of ±100 µm. However, if severe soldering problems are to be avoided, higher placement accuracy will be required for very small components or devices with fine pitch lead spacing. Pick-and-place equipment with an improved placement accuracy of ±0.05 mm (2 mil), and a repeatability of 0.02 mm (0.8 mil) is available, but generally has a lower placement rate.
Common sources of inaccuracy for X, Y, and q include:
The next sections cover the main areas in which inaccuracies affect component placement. Then, because inaccurate placement adversely affects yield and increases rework, we will look at methods of error correction which are an essential element in design for manufacturability.
Note that a small amount of manual adjustment after automated placement can reduce the amount of rework, but it can also damage components. If a process regularly requires manual adjustment of components after placement, it is better to analyse the factors affecting placement accuracy and make appropriate amendments to the process to remove the problem.
Machine inaccuracy is attributable both to control and measurement uncertainties and inaccuracies which centre around the vision system, and to a range of mechanical factors which produce translational (X, Y) deviations and rotational (q) inaccuracy, specifically the dynamic behaviour and reproducibility of the drives, the machine stability and heat effects.
Any dimensional variation, even between boards from different suppliers, should be within the manufacturing specification, provided that they are working to the same revision of the artwork! However, the manufacturing accuracy achieved in a printed circuit board will depend on factors such as:
For FR-4 material, 300 mm long circuits manufactured at a temperature of say 15°C, will be 50 µm longer if the assembly area ambient temperature is 25°C. This factor alone could be sufficient for out-of-tolerance placement on fine-pitch devices.
Also multilayer boards may warp because of the way they are stacked or laminated. More seriously, this warping may be uneven, due to variations in the density of the copper patterns.
The track pattern on the printed board may also have errors relative to the fiducials on the board. Examples of errors are:
Track patterns can however generally be made within 30–40 µm of intended position in relation to the fiducials. Where greater accuracy is required for a hard-to-place device, this is normally achieved by using a fiducial local to the device as well as the overall board fiducials.
Printed circuit boards containing both leaded and surface mount components are generally located using two precision reference holes near the long edge and spaced as far apart as the board will allow. When automatic board handling is not used, the relationship of the locating holes to the outside board edges is not highly critical. On the other hand, if automatic board handling is also a requirement, this relationship becomes significantly more important, as the board transport handling is accomplished by pushing and locating on the board edges.
Once the board is roughly in position, the locating holes must be within the ‘pull-in range’ of the pin locators. This type of location results in a repeatable final location generally within ±0.025 mm.
For assemblies containing only surface mount components there may not be any need for any holes in the board. In such a case, edge location must be used. The only caveat about the use of board edges as precision locators is that the same locating points on the edges must be used for all operations in order to allow for the edges not being exactly even or straight. Usually two locations on the long side are used in conjunction with an edge location near the reference side. The finish of the edge must be reasonably smooth so as not to affect the location mechanism, that is, not scored or broken.
A component has a certain amount of inherent dimensional inaccuracy, which is seen as variations in:
Although the resulting parts are all within the manufacturing specification, there will be variations both between components from different suppliers and from batch to batch from a single source, all of which may require placement machine adjustments.
In particular, size tolerances of ceramic parts, such as multilayer capacitors, are rather large, which is attributable to the pressing and sintering processes during component manufacture: both process aim-points and spread of results will vary between suppliers. The situation gets worse as the packages get smaller, because not all quantities scale down in proportion to the overall size: for example, the thickness of component metallisation remains unchanged.
Table 1 gives manufacturing dimensions for a range of common chip component sizes, showing some significant variations:
Fortunately, passive devices are generally small in overall dimension and are mounted on pads which are generously proportioned. For these reasons, problems associated with manufacturing dimensional accuracy do not generally present any cause for concern. Any problems tend to be due to variations in thickness between manufacturers, which can result in differences in placement pressure.
However, integrated circuits present very real potential for possible deviations due to manufacturing dimensional inaccuracy. As an example to show these tolerances, Table 2 gives data for two Quad Flat Pack IC packages, from which it can be seen that lead length can vary by as much as 0.30 mm, and lead width by 0.20 mm.
Pitch = 0.8
Pitch = 0.5
With fine-pitch lead spacing, the width of each footprint is half the spacing. Fine-pitch spacing below 0.5 mm (20 mil) and ultra-fine-pitch with 0.25 mm (10 mil) have become real requirements. This means footprints 0.125 mm (5 mil) wide, and lateral placement accuracy of ±0.06 mm (2.5 mil).
Angular accuracy of placement is also a major consideration. For a QFP with dimensions of 25 mm × 25 mm, an angle of twist of 1° means a lateral displacement of 0.22 mm (8 mil) at every corner. As a result, with a fine-pitch layout, about half the leads would sit on the wrong pad. Therefore with large fine-pitch components, the angular placement accuracy has to move towards ±0.1°.
For non-critical circuits, alignment of the board in the placement machine may be achieved by referencing placement direct to the jigging holes, but for most boards special etched marks, called fiducial marks or ‘fiducials’, are used for automatic alignment, with vision systems incorporated within the placement machine.
Component vision systems are also used to correct for inaccuracies in position of the picked components, replacing the previous mechanical centring system, whose use is now restricted to a handful of suppliers. The term ‘optical centring’ is sometimes used, although no physical centring occurs. Instead, the vision system calculates the position and placement angle of the component after pick-up by the vacuum nozzle, and these are used to adjust the placement program.
For small devices and fine-pitch devices with large pin counts, fiducial alignment of both PCB and component is advised. For devices with lead pitches below 0.4 mm, fiducial marks close to the component footprint should be used to compensate for local inaccuracies and non-linear warping of the PCB.
The accuracy of location of a fiducial mark by an optical vision system can be as close as 20 µm. However, this will vary with the chosen system parameters such as field-of-view, the optical system ‘gain’ and magnification characteristics, the illumination system, and the algorithm used for image analysis. There are different styles of fiducial mark which are best for particular combinations, and it is always wise to check with the manufacturer of the placement equipment to find out what is the preferred size and shape of fiducial for their system. Other concerns are the colour and/or contrast of the fiducial compared to the circuit medium and the possibility of confusing the fiducial with nearby pattern elements.
When using multi-fiducials, rotational errors can generally be corrected to within 0.2°, depending on the machine capability. In analysing equipment performance, however, remember that, despite the specification of the machine, there may be disparities between:
After each board has been loaded onto the machine, a camera (usually mounted on the placement head) sequentially records the position of the fiducial marks and calculates the board’s position and placement angle compared with its expected position relative to the machine’s coordinate system or ‘frame of reference’. All placement positions and angles are then corrected by applying this board-specific offset to the placement coordinates for each component, a process referred to as coordinate transformation.
The algorithm used will depend on the preferences of the system designer and will therefore vary from machine to machine. A straightforward, but comparatively slow approach is the two-part correction described below.
In Stage 1, the position of the bottom left fiducial of the board being processed is compared with its expected position relative to the machine frame of reference. The system then moves the board first in X and then in Y direction until the vision system detects that the board fiducial has aligned with the machine reference. The X and Y offsets are then calculated as deviations for the ideal position and the results stored in a log file for later use (Figure 4).
Stage 2 considers the fiducial mark which is diagonally opposite. A line can be drawn between each of the two sets of fiducial marks: since one of each pair has been aligned and the other not, the lines will show the rotational offset between the board and the machine coordinates.
Once again the frame of reference is moved until it overlays the fiducial mark. The movement in this case is rotation about the first fiducial position. Once the fiducials have been aligned, the angular rotation needed is stored in a log file (Figure 5).
The board offset coordinates, X and Y together with the angular rotation q, can then be used as a global board offset for the placement coordinates of all of the devices to be placed by the machine on the board. Thus the machine compensates automatically for the variation in board loading position.
The position of fiducial marks is thus measured and overall correction applied to the whole programme. This confines remaining placement errors to copper accuracy, component outline, component centring and accuracy of placement.
A centring tweezer was the original method for ensuring the correct position and orientation of components relative to the nozzle, and this could be done at pick-up, ‘on the fly’, or at a separate station. However, mechanically moving a component on the nozzle can lead to damaged components and is best suited to components with regular outlines.
In the past, the use of vision systems for centring was restricted primarily to large fine pitch modules, where they were required for accuracy reasons: even a very small rotation can produce an unacceptable result. However, with the advent of cheap and very powerful computers for image analysis, vision systems have almost entirely replaced mechanical centring2 systems for all sizes of component. The term ‘optical centring’ is sometimes used, but no physical centring occurs.
2 Some machines (for example, the Mydata TP9) are still found with a mechanical system, but now these are ‘soft touch’, making several contacts with body and lead tips in order to centre the part, rather than the original tweezer grippers which operated simultaneously in X and Y.
Compared with mechanical centring, vision systems give:
Vision systems check each component after pick-up by the vacuum nozzle, and an individual correction for position and angle is then made to the placement programme. The images that are processed in order to calculate the position of the component relative to the centre of movement of the nozzle are taken by ‘parts cameras’, which are usually fixed3 to the framework of the machine.
3 Parts cameras can also be movable, for example, attached to the placement head. However, the fixed camera option is recommended as removing a potential cause of error.
The ways in which images are processed vary considerably in sophistication and detail. For the simplest parts, just the centre of the body and the outline of the part may be derived from the information collected; for more complex parts, the central position of each lead may also be estimated, and the placement position adjusted so as to get the best statistical fit between part and board, bearing in mind the variability of the measurements.
Vision systems are used during placement for a range of functions, which include:
The last section has dealt with the first three of these. In this section we will be looking at some of the practical issues concerning vision systems, particularly for the component camera4, and giving some information on the other camera uses.
4 The description that follows describes an optical alignment system where an direct image of the component and its actual position is compared against the expectations built-in by the programmer. Be aware that not all alignment systems work this way – many systems work with multiple camera images to overcome the problem of simultaneously getting sufficient resolution and a view of the overall package, and more radical alternatives include laser systems.
A typical monochrome camera has a CCD (charge coupled device) photo sensitive chip which splits the image into 512 × 512 pixels. Like the human eye, the camera produces a signal proportional to the strength of light received, but typically digitises the information, for example, outputting 256 levels (coded 0–255).
In many older placement machines a binary conversion technique is used to increase the speed at which the computer processes this signal. The conversion divides the signal into two levels (0 for bright and 1 for dark) depending on whether the brightness falls above or below a programmed level. This is referred to as the CCD ‘threshold level’, and its value has a great influence on the overall effectiveness of the vision processing system. Only if the CCD threshold level is set correctly can the vision processing system accurately read symbols such as fiducials and block skip marks.
An important requirement for any optical system is good illumination. Fiducials on boards always need to be viewed by reflected light, but components being mounted can be illuminated in two ways:
Back lighting has been very popular, being more accurate and more compatible with binary conversion systems. However it is limited to devices with protruding leads. The illumination coming from behind the device provides a good silhouette of the leads, but can be ‘misled’ by moulding flash on plastic components.
If the leads are on the bottom of the body, as with PLCCs, reflected light must be used. In this second method, LEDs generally illuminate the device so that the leads are highlighted more than the ‘darker’ body.
With continual increases in computer power it has become possible to process more signal information in this ‘real-time’ situation, and there is a growing trend to use all 256 levels of brightness in the so-called ‘grey-scale’ vision systems. These are capable of making better judgements about position because it is possible to view the underside of the component and differentiate, for example, between leads and body in the case of a J-lead package. Grey-scale systems are usually associated with under-body illumination.
There are several formats of fiducial marks currently in use (Figure 6), and it is always wise to check with the assembler so that any requirements for a particular fiducial format can be incorporated into the artwork. It has been known for a customer to include up to three different fiducial formats on boards in order to be in a position to be able to move boards between assemblers!
The recommendation from SMEMA, the trade association for machine manufacturers, is that:
Although some devices are physically symmetrical, they must still be placed in a particular orientation on the board to ensure that the polarity is correct. Machines operate on the premise that all the parts from a feeder are orientated in the same direction when they are picked, but in some cases a part may be oriented differently, through error by the supplier or operator.
It is therefore desirable to check the actual orientation of each part that is picked, either by having the parts camera search for a cut or a notch on the part, or by having an separate 1st pin camera check the position of letters or symbols printed on the surface of the device. This function is known as ‘1st pin check’.
As the solder paste layer is thin, even a small amount of non-coplanarity on the component leads can produce open-circuit joints. To minimise the number of ‘bent pins’ mounted, the coplanarity of all leads on a module may be measured before placement, although this increases equipment cost and adds 4–5s to the placement time. One of the methods which has been developed uses the following sequence:
The limiting value for coplanarity will depend on solder thickness, lead pitch, and the thickness and possible camber of the PCB: 100 µm is typical for 0.65 mm pitch leads; 75 µm for 0.5 mm pitch.
What are the sources of placement inaccuracy that will be important for a 0603 capacitor? How will these differ for a QFP-208? And how should the board designer allow for the correction of inaccuracies?
From the term ‘pick-and-place’ one might think the mechanics of the process to be quite basic, and indeed a high standard of hand placement is possible, given a careful operator and enough time. However, a modern placement machine head is a highly complex and sophisticated piece of equipment.
The component placement head (Figure 7) must carry out the functions of:
A number of sensors must also be fitted to the head to ensure correct machine operation.
Depending on the design and application, the head may also be involved in:
5 If the assembly is to be wave soldered, the placement operation needs to be preceded by glue application, to deposit a metered amount of adhesive between the component footprints. This task can be carried out either on the placement machine or using a dedicated glue station.
The key features of the resulting placement system are individually reviewed in the sections which follow.
The surface mount devices are held by a vacuum nozzle, also known as a ‘pipette’, whose size is determined by the size, mass, and centre of gravity of the device. A nozzle has no inherent centring action, and will merely pick up a component whether or not it is central, as long as the nozzle makes contact with its face.
Rotation of the nozzle is important for two reasons:
In theory this rotation could be provided by table movement, but combining X, Y and q movement in one stage is complex and inefficient.
Dual-heads enable a machine to place a component with one head, while picking up the next component with the other head. In other configurations, one head may place chip attach adhesive while another is placing components on the board to which adhesive has just been applied.
In many machines the placement head is fixed to the carriage of an X/Y manipulator driven by servo-motors. In others, the head is fixed and the board moves. Maximum speeds are of the order of 1.5m/s, with maximum accelerations of about 15m/s2. The speed and acceleration of the alignment mechanism directly affect the component placement cycle time or ‘takt time’. This is typically 0.1 seconds/component for standard chips and using a single head.
The nozzle also has to place the component with sufficiently downward pressure that the ‘tack’ in the paste or glue will retain it in position. Typically, controlled forces in the range 2–8N are needed to give a consistent result.
With high speed placers, the pressure comes from the resistance of the board to a preset degree of over-travel of the nozzle, calculated from the nominal thickness of component and board. As the designed over-travel is typically only of the order of 25–50 µm, the actual pressure can vary considerably if component thickness varies between batches.
With larger components, and with more sophisticated machines, the small ‘Z-force’ required may be applied by a servo motor, with a programmable placement force controlled by feedback from a transducer.
Theta alignment of the component is provided by rotating the vacuum nozzle. Both coarse (45/90° steps) and fine adjustment is needed: precision machines offer programming in increments as small as 0.01°. The rotation is produced by a servo-driven mechanism.
Particularly where high accuracy is required, the coarse rotation step may precede vision checking, in order to prevent any errors being introduced by rapid coarse rotation.
The heads on typical ‘chip-shooters’ carry a small number of nozzles which are selected by rotating the head to present the nozzle which is correct for the size of the next component to be placed. Slower machines with an automatic selection facility take nozzles from a tool-bank situated at the side or back of the machine. This allows a much wider selection of nozzles to be used, but remember that every tool change takes time!
The board transport module performs the following functions:
The feeding and removal of boards to and from the working area is mostly carried out on conveyor belts. The Surface Mount Equipment Manufacturers Association recommend that support edges of 4.75 mm are provided on both sides of the board (Figure 8). In order to provide satisfactory support of the board on the conveyor, no components may be located on these edges.
Typical speeds for run-in and run-out are 300–400 mm/s. The board is stopped at the required position by means of a mechanical stop or an optical sensor. The latter has the advantage that the PCB does not receive any impact as the result of colliding with a mechanical stop.
The board can be positioned (Figure 9) in the placement area in three ways, the third of which is now almost universally adopted:
The board is usually supported on the underside by support pins, to ensure that the board does not bend too much due to the placement forces. The height of these pins is determined by the height of any components mounted on the underside, and is generally between 15 mm and 50 mm.
Most modern placement systems achieve high placement efficiencies through the strategies of:
Although there are many variants, and some high-speed dedicated machines that use other principles, the configuration of most general-purpose pick-and-place systems conforms to one of four ways of implementing these strategies:
Comment: optimum flexibility as regards component and feeder types, but relatively slow because of time spent in head movement.
Comment: optimised for speed, but limited to relatively small components.
Comment: fastest solution, but limited in flexibility – optimisation of placement sequence is essential in order to use all placement heads properly.
There are also some hybrids of the last two types. However, it is the first two that are the ‘main runners’, so these are described in more detail in the following sections.
The original concept of picking and placing one component at a time was not able to keep pace with demands for placing high volumes of parts, so separate specialised systems gradually developed for placing high-pin-count modules and fine pitch components, whilst placement of chip components and low-pin-count components was taken on by ‘chip-shooters’. This process is continually being refined to achieve as high as possible a placement rate whilst maintaining full flexibility.
The current chip-shooter is the result of a number of different design approaches where heads are grouped in a ring, and proceed progressively to pick up components from feeder positions and deliver them to a common placement area, to which the board is taken. Some early designs had fixed feeders, distributed around a circle, with a correspondingly large ‘turret’ of many placement heads. This had an advantage, in that feeders could be replaced during operation, but the increased size, complexity and number of heads has driven the chip-shooter approach to that described in Figure 10.
As you will see from the photographs, a typical chip-shooter has a relatively small turret, combined with extensive banks of feeders which move (very fast) along a track at the rear of the machine Most systems have dual feeder banks, so that one can be operated while the other is reloaded or replenished. Parts are picked up from one of the feeder banks by the nozzle on the head at the rear of the machine, and then carried progressively through a series of operations, which includes coarse rotation, component recognition, fine rotation and then placement, as shown in Figure 11.
The operation of each of the heads on the turret is quite complex, made more so by the need for each head to be fitted with a number of different nozzles in order to accommodate different sizes of device. Originally equipment had two or three options, with nozzles selected either coaxially or by having several nozzles close together, moved mechanically into position. The need for a greater variety of nozzles to accommodate an increased variety of components has meant that newer designs frequently have up to six nozzles fitted, often in the star wheel format we saw earlier – those of you with different backgrounds might prefer to think of this as a spur rowel or a leather punch!
Specifically mentioning that nozzles are moved mechanically does not imply that other elements are necessarily electrically or pneumatically driven. In fact, the nature of the rotating head means that most designers have found it convenient to control the operation of individual elements of each head by means of cams on the machine, the cam follower path being linked to the requisite motion. In consequence, the turrets are complex and need substantial maintenance to ensure continued satisfactory performance.
Rotation is generally applied through gearing from a motor at the appropriate turret position which is brought momentarily into contact with the head, although this is not the only solution.
As the main up and down movement to pick up and place is generally cam controlled, it is quite usual for machines to be set so that they would normally press the nozzle 25 µm below the surface of the board, but programme the board support mechanism to move downwards by the expected height of the component. In practice, because most chip-shooters are operating on components of very similar thickness, you will not see substantial movement.
Whilst there are 12, 16, or 20 heads, and correspondingly the same number of positions around the circle, not every position is used, partly for mechanical reasons. In almost all machines, half the circle is given up to checking that the component has been cleared from the nozzle, restoring the initial setting, and readying the correct nozzle for the next component to be picked up.
Some indication of how the system works is shown in the linked video clip, CSPB.Mov. but don’t attempt to run this unless you have a reasonably fast web connection. You will see first the placement of a series of components close together (a typical way of demonstrating placement accuracy) and the red flashing LED illumination of components used for component recognition.
Whilst turrets have a full set of heads, the assembler is not always able to make optimum use of these. In some cases, certain heads may be dedicated to have special-purpose nozzles, such as those used for MELF components; in others, the head will have been disabled automatically by the machine control system because of nozzle damage or other problems which led to the head being identified as potentially defective.
Over all, the chip-shooter is a piece of equipment that is worth careful maintenance, and when properly working is able to achieve placement rates of 30–40,000 components/hour. Depending on the machine, it can also handle a considerable variety of components. For example, many machines have twin feeder banks, each capable of holding 70 feeders. Be aware however, that the size of the banks is always measured in terms of 8 mm feeders, and larger components take 2, 3 or even more slots. If you have a wide variety of components to place, then it is worth considering whether it is possible to mount these in a single pass – exceeding what is an absolute limit by only one component type can make a substantial impact on cost, and this is another reason for trying to reduce to a minimum the number of different components used on a circuit.
A precision placement machine typically picks up components from a bank of feeders, either singly or a few components at a time, then moves them over a light box where the ‘snapshot’ of the component position is taken, then finally to the placement area on the board. This is the originally pick-and-place process, except that vision systems have replaced the previous mechanical centring, which took place either on the head or at a fixed centring station.
Precision placement machines vary enormously in their organisation, smaller machines frequently allowing feeders to be placed on all four sides of the machine, with waffle trays of components competing with the board for space in the bed of the machine. In these the placement head does all the work.
The reason for using all four sides is that feeders for large components are wide, and take up a lot of space. However, when one tries to integrate a placement machine into a production line, allowance has to be made for conveyors to carry boards in and out of the machine. In consequence, most production machines, such as those shown in the photographs, have feeders only at the front and rear, to allow operator access.
With a chip-shooter, it is not uncommon for the feeder to be incremented mechanically as a nozzle descends to pick up the component, by depressing a lever with another part of the head; with precision placers, the connection to feeders is normally electrical or pneumatic. Another difference is that the feeders are static, which allows the operator to splice on new reels of components ‘on the fly’, rather than halt the machine (though safety considerations sometimes preclude this).
A final difference is the way in which tape waste is dealt with – on the chip-shooter, there will always be some mechanism for cutting off small sections of the bottom layer tape, rather than letting them project into the machine; with precision placers, proper facilities can be available for collecting the waste tape without generating cuttings or dust.
There are a number of ways of organising the moving head: on the version seen the head gantry allows the head (one of two) to move in both X and Y directions; on other machines the head movement is restricted to one axis, traversing the pick-up position, light box and placement position only along the X axis, with the nozzle bank and board able to move in the Y axis. This latter approach was used by Fuji in their IP series of machines. However, whilst simplifying the mechanics, it substantially slowed down operation, and a more common approach is slowed down operation, and a more common approach is to have an independent head.
In all precision placement machines, great care is taken to have the right quality vision system, with lighting from below, and (usually) a reflective underside on the nozzle surround to ensure that the part is back-lit. In some cases, the part is brought to a stop above the light box, the ‘snapshot’ taken, and the part then moved to the placement position. Other systems use flashes of either LED or Xenon light to ‘freeze’ the image of the component in mid-flight. These need high-intensity light, and sensitive CCD cameras, but generally result in faster placement.
Precision placers typically operate in the region of 2,500 to 8,000 components/hour, depending partly on component size. One element that certainly slows up any machine, is using components from trays. In order to increase productivity, it is usual to have a waffle-tray feeder bank on the side of the machine, with separate mechanisms to locate components and present them to the placement head.
Other complications in precision placers that can slow up the placement rate include first pin inspection and (especially) coplanarity testing – only specify this if it is absolutely vital, and you can control the planarity of the part in no other way. However, it is well worth looking at the capability of your vision system for components such as BGAs, where eliminating devices with missing balls saves costly rework.
Visit your assembly line (or go there in your memory!) and take a good look at the different types of placement equipment. Try and collect answers to the following questions:
This is an activity to which there are no correct answers!
You don’t need to spend too much time on the detail of this activity during your study of this module, but keep in mind as you visit assembly plants during your work that you can learn a lot from watching real machines operate. Think about how their designers have approached the trade-offs between speed and accuracy, and the challenge of making sufficient types of component accessible.