Known methods and devices for automated placement use electronic image recognition methods for recognising, in an image of a supply of circuit components, an individual component, to determine its location and orientation and to guide a manipulating device to the location of the circuit component and to seize it. As manipulating devices, vacuum suction tips are widely used, which may be supplied with a negative pressure in order to suck the component to be seized and to hold it until it is placed at the intended location of the circuit carrier by releasing the negative pressure. By measuring the negative pressure at the suction tip, it may be found out whether a circuit has been seized successfully or not. If the seizure has failed, the suction tip is open, and the negative pressure present there is noticeably less than when a successfully seized component covers the opening of the suction tip.

There are types of circuit components which are difficult to manipulate by this prior art technique. Among these, there are the so-called beamlead components. These are semiconductor components having a very small chip size of typically about 100 to 400 , um edge length, which carry laterally protruding flat metallic connecting or soldering vanes for making contacts. Such components are difficult to manipulate due to their small size, the brittleness of the semiconductor material and the sensitivity of the

connection vane, but in particular in some high frequency applications they are indispensable because they achieve extremely low parasitic capacities and, hence, high switching speeds.

When attempting to automate the assembly of such components, a number of difficulties arise. On the one hand, the open cross-section of suction tip to be used for seizing the components must be very small, so that it may be placed completely on the semiconductor crystal of the component and that there is no danger of the sucked component tilting and the semiconductor crystal being damaged. In order to protect the sensitive components, the suction force must be small. Due to these constraints, it is not possible to judge with certainty whether a component has been seized successfully or not by measuring the pressure at the suction tip. If the seizure of the component fails and this is not recognized, there is a danger that in a placing attempt on the circuit carrier, instead of the component believed to be held at the suction tip, the suction tip touches an adhesive that was applied to the circuit carrier earlier. Then the placing procedure must be interrupted in order to clean the suction tip.

Another problem in the automated manipulation of beamlead elements is the correct recognition of their orientation. Generally it is not difficult to automatically recognize the body of a beamlead component before a background, the brightness of which differs sufficiently from that of the body. However, the body generally has no features from which the orientation of the component might be concluded. The orientation of a beamlead component is usually only recognisable from the shape of its contact vanes, one of which conventionally has a pointed end and the other one has a swallowtail-

shaped one. It is possible to recognize these contact vanes automatically in an image of the component taken before a dark background, but the image processing methods conventionally used for automatic placement are not capable of recognizing the two contact vanes as parts of a same object.

The object of the present invention is to provide a method for automated placement of circuit components on a circuit carrier, which is suitable for manipulating highly sensitive components, in particular beamlead components.

The object is achieved by a method having the features of claim 1.

Since the presence of the component at the manipulating device is verified optically, there are no limitations as to an appropriate technique for seizing the components.

This may be done conventionally by sucking; since no detectable pressure drop is needed for recognizing the presence of the component at the manipulating device, the suction power may be set at a low value, so that it is sufficient for seizing the circuit component, but that there is no danger of damaging the component by the suction force.

In order to determine the location of a component to be placed before it is seized by the manipulating device, it is preferred to use a camera which is displaceable along with the manipulating device and which may also serve to verify the placement of the component on the circuit carrier when it has been deposited there by the manipulating device.

For checking the presence of the component at the manipulating device, another camera is preferably used, which may be stationary. There are already automated placing systems having two such cameras, wherein the first one conventionally is used for locating electric contact points at the bottom side of a flip-chip component held at the manipulating device. Such a placement system can therefore be adapted for carrying out the present invention without substantial apparatus additions, merely by adapting its operating program.

In order to place the manipulating device exactly on the body of the sucked component while seizing it, even if this body is not immediately recognisable in an image generated by the first camera, it is preferred first to determine in this image the locations of at least two regions of the component, in particular of its contact vanes, and the location of the body is calculated from the locations of the regions according to a model of the component. In the simplet case, this model may be just be formed of the information that the body is to be found in the centre between the two contact vanes.

If the two regions have different shapes, the orientation of the circuit component may be detected from the image taken by the first camera by distinguishing between these regions.

A preferred form of manipulating device for manipulating beamlead components has a suction tip for positioning above the body of the beamlead component and feet

connected to the suction tip for placing above the contact vanes. Feet and suction tip are placed with respect to each other so that when the feet touch the contact vanes, a gap remains between the suction tip and the body, and the negative pressure supplied to the suction tip is set so that even when the component is seized by the manipulating device, a space remains between the suction tip and the body. Damage to the body by the suction tip is thus excluded.

For verifying the presence of the seized component, it is possible to search in the image taken by the second camera for the outline of the component in a way known as such. If the component is dark, the manipulating device must form a bright background in order for the outline of the component to be easily recognizable. In front of this bright background, a suction opening of the suction tip will also form a dark image, if the component is not present. In order to prevent errors, it is therefore more appropriate to search for bright regions of the component in an image taken before a dark background, and to consider the component as seized successfully if at least one such region is identified.

Preferably two such regions are identified, and if these two regions are the contact vanes of a beamlead component, the orientation of the component may be detected based on this identification, unless this has been done before already based on the image provided by the first camera before seizing the component, or the orientation already recognized at that instant may be verified once more based on the image taken by the second camera.

Further features and advantages of the invention become apparent from the subsequent description of embodiments thereof referring to the appended drawings.

Fig. 1 is a schematic view of a beamlead element in a so-called waffle pack ; Fig. 2 is a schematic representation of a manipulating device and a first camera for seizing the beamlead component from the waffle pack; Fig. 3 is a schematic representation of the manipulating device with the seized beamlead element in front of a second camera; Fig. 4 is an image taken by the second camera; and Fig. 5 shows the manipulating device while placing the beamlead element on a circuit carrier.

Fig. 1 is a schematic view of a beamlead component 1 having a body 2 made of a structured semiconductor crystal and two contact vanes 3,4 that extend from the body 2 in opposite directions and are formed of flat metals strips, one 3 of which is pointed at its free end and the other 4 is swallowtail-shaped. Such components are frequently distributed by the manufacturers in so called waffle packs, plastic trays in which a plurality of chambers 5 is stamped, each of which receives a component. When not in use, the chambers 5 are closed by a common lid, which is removed before taking out the components and which is therefore not shown in the figure. Since the chamber 5 is

much larger than the component 1 and the component 1 is freely movable in chamber 5, the exact location of the component I must be determined before the component 1 can be seized automatically by a manipulating device.

Fig. 2 shows a manipulating device 6 and an electronic camera 7, which are movably mounted together to a robot arm or a similar tool and which are positioned above a chamber 5 in order to seize a beamlead element 1 in it.

The manipulating device 6, shown partly in section, has an essentially conical body through which a suction channel 8 extends axially. As can be seen in the bottom view of the manipulating device in Fig. 3, the lower mouth of the suction channel 8 is surrounded by an annular flat seat 9. Two feet 10 are formed diametrally at the flanks of the cone. The level difference between the soles 11 of the feet and the seat 9 is slightly more than the height of the body 2 of the component 1.

A microprocessor, not shown, receives an image of the chamber 5 recorded by camera 7 and searches this image for the outlines of the contact vanes 3,4 of the component 1. In order to facilitate their recognition, the background of the image is made dark, for example, by using a waffle pack of a dark material, or by placing a transparent waffle pack on a dark support. From the position of the contact vanes 3,4 in the image, the microprocessor calculates their positions in space and that of the body 2 between them. Based on this information, the manipulating device 6 is positioned exactly above the component 1 and is then descended, so that its feet 10 come to rest on the contact vanes 3,4. Since the level difference between the soles 11 and the seat

9 is more than the height of the body 2, the seat 9 does not come into direct contact with the body 2.

By distinguishing between the contact vanes 3,4 in the image of camera 7, the microprocessor is capable of detecting the orientation of the component 1 and to take it into account later, when placing it on a circuit carrier.

The suction channel 8 is now supplied with a negative pressure which is selected so that it is sufficient for raising the component 1, but not for bending the contact vanes 3,4 so strongly that the body 2 comes into contact with the seat 9.

Fig. 4 shows the manipulating device with the thus seized component 1 in a position above a second camera 12, which is stationary and which is visited by the manipulating device 6 on its way from the chamber 5 to a circuit carrier to be equipped, in order to verify that the ; component 1 has indeed been successfully seized.

Fig. 5 schematically shows an image as is recorded by the camera 12 if the component 1 was seized successfully, in contrast to the image of Fig. 3 which would be recorded by the camera 12 if the seizure of the component had failed. While the body 2 of the component 1 is dark, like the body of the manipulating device 6, so that there is little contrast between the two and they are not distinguishable from one another with certainty in the image, the bright reflecting contact vanes 3,4 show clearly. For identifying these in the image, essentially the same program routines may be used as before in the image generated by camera 7. If the two contact vanes 3,4 are

successfully recognized in the image of camera 12, the component 1 was seized correctly, and the manipulating device is moved on to a circuit carrier 13 in order to place there, as shown in Fig. 6, the component 1 with its contact vanes 3,4 on a dot 14 of an electrically conductive adhesive previously applied to the circuit carrier 13.

If the identification of the contact vanes 3,4 in the image recorded by camera 12 fails, it is assumed that the component 1 was not seized correctly, and the manipulating device 6 and the camera 7 are moved back to the chamber 5 in order to check whether the component 1 has remained therein. If confirmed, it is attempted once more to seize it, otherwise, there are moved to another chamber 5 of the waffle pack and a component is seized from there.

Using the above described method, highly sensitive components may be automatically placed reliably and precisely on a circuit carrier. Failed seizures are recognized and aborted before they can cause malfunctions that must be corrected by human intervention. Since damage to the components during placing is practically excluded, nearly all the obtained assembled circuits are operational; there is almost no scrap or repair of defective circuits.