FI104290B - Method for testing the reliability of integrated circuits - Google Patents

Description

104290

This invention relates to a method for testing the reliability of integrated circuits. More particularly, the invention relates to testing methods in which the connections in each of the 5 circuits under test are coupled as desired to provide a suitable resistor in the circuit for testing purposes, and the test circuits are connected to a power supply to supply a measurement current through the circuits to be tested.

New low-volume component fabrication technologies, where wiring connections 10 are directly applied to silicon substrates such as flip chip, COB (Chip on Board), CSP (Chip Scale Packaging) and TAP (Tape Automated Bonding), are of interest, especially in small and portable but increasingly complex electronic products. . The reliability of these types of circuits is key to reliability, so the manufacturer must ensure the reliability of the component technology connections he chooses before moving on to them.

15 One of the most common testing methods is to expose the circuits to cyclical temperature variations. This is done by connecting the circuits of the circuits together in a so-called. daisy-chain configuration by placing the number of test circuits required for statistically significant results in a test cabinet and supplying them with a current that causes a voltage drop across the connections. This voltage drop is monitored either directly over the circuit junctions or by measuring over the exact 20 cross-section resistors in series with Senja, whereby the change in junctions causes a change in current and is thus reflected as a voltage change in the meter.

The problem is that each circuit must have its own measuring channel, or at least its own power supply, whereby the amount of cabling for the power supply is 2n, where n = the number of components to be tested. If you want to use the so-called. In addition to the four-point method, each component is additionally equipped with data acquisition cables, resulting in a total cabling of 4n. Thus, parallel power supply increases the number of channels and cabling. According to one calculation, there must be at least 32 identical samples to be tested if the results are to be presented statistically. This problem is exacerbated when it comes to comparing components from many different manufacturers, as well as connections made using several different technologies and process parameters. For example, testing joints made by 10 different processes would require 320 channels at the minimum sample size, which would be prohibitively expensive.

2 104290

On the other hand, successive testing of different processes on the same equipment is not an issue when you remember that the normal test time is 1000 h, ie about 1.5 months.

Another problem is that parallel supply will never provide absolutely equal current to all components, because of the connections, cable lengths, 5 possible applied resistance, etc., there will always be uncontrolled voltage losses, which will change the magnitude of the current. .

Of course, the circuits can also be connected in series, reducing the need for cabling. The problem here is that if one of the circuits connected in series breaks, all circuits 10 will remain unpowered and will not be tested for them. Thus, in known methods, the circuits are connected in parallel, and each of the chained circuits is provided with its own power supply. Applicant's earlier application FI 970378 discloses a method based on switching a diode downstream of each of the circuits under test and supplying the measuring current from a constant current source such that a break in the circuit caused by a broken circuit causes the diode 15 to go into a conductive state. Although this method allows the test to continue automatically despite disconnection, it does not provide a solution for the immediate identification of a broken circuit, or for monitoring other phenomena seen in the circuit voltage curves.

It is an object of the present invention to provide a method for testing the reliability of integrated circuits, which avoids the disadvantages of the above methods, and by which the measuring channels of existing measuring devices can be simply multiplied. Various embodiments of the invention are characterized by what is set forth in the claims below. In the measuring method according to the invention, the operation of the circuit is based on the threshold voltage of the daisy-chain, that is, the diodes connected over the series-connected circuits. By switching a different number, e.g., a growing number of diodes in the direction of the chain over each circuit, each measuring circuit has its own measuring range. By conducting a constant current through the components and observing the voltage across the circuits in the circuit, the value of the voltage can be seen if any of the components in the diode circuit in question is damaged or its resistance is changed. By means of the invention, several components can be connected to the channels of existing measuring devices for continuous monitoring, and information on voltage values, e.g., fed into a spreadsheet program, can simply detect which component has been broken or which component has changed resistance.

3, 104290

The invention will now be described in more detail by way of example with reference to the accompanying drawings, in which Figure 1 illustrates a way of connecting so-called circuit connections. a daisy-chain method; Fig. 2a illustrates the connection principle of the method according to the invention in the so-called circuits.

5 daisy-chain coupling;

Fig. 2b shows voltage curves corresponding to Fig. 2a; Fig. 3a shows a measuring circuit corresponding to the method according to the invention; Fig. 3b shows a voltage diagram corresponding to Fig. 3a.

Figure 1 illustrates the daisy-chain principle, in which all the connecting bumps 1 10 of the circuit are interconnected so that they are connected to each other alternately on the substrate side 2a and on the circuit side 2b. This results in a relatively high resistance value between the power supply points of the circuit. 1. with a four-point connection, the measuring current I is supplied to the circuit by a power supply 3 along different wires and at positions la over which the voltage across the circuit (mV) is measured by a voltage meter 4. Thus, the voltage losses of the power lines and terminals do not affect the daisy-15 circuit. In the known parallel current supply solutions, four wires are required for each circuit, and each circuit occupies one measuring channel.

Fig. 2a shows a coupling principle corresponding to the method according to the invention, in which the circuits are replaced by 20 ohm resistors R1 and R2. One diode D is connected across the first circuit, i.e. resistor R1, and two diodes D are connected across resistor R2. The diodes D (e.g., 1N5626) are connected in a direction parallel to the measuring current I. The measuring current I from a constant current source (not shown) is of a magnitude (e.g. 10 mA) such that the voltage across resistors R1 and R2 is lower, e.g., about 0.2 V, than the threshold voltage of a diode, e.g. 0,63V / 10mA. When a measuring current I (10 mA) is applied to the earth-connected resistors R1 and R2, and one of them fails, the simulation results according to Fig. 2b are obtained with a voltage meter V. In Fig. 2b, the lower voltage curve a is associated with a failure of a circuit represented by a resistor R1 and an upper b is a fault with a circuit represented by a resistor R2. From the curve it can be stated that the voltage level of the lower curve (failure of circuit R1) never exceeds 0.62 V with given parameters, so all higher voltage levels must be related to failure of circuit R2. The smallest resistance value of the circuits with which the cases can be reliably distinguished is on the horizontal axis describing their resistance values at about 30 to 45 ohms. The voltages in tabular form can be represented as follows:

Limit 45 ohm to 100 ohm

1,543 mV 622 mV

2,640 mV 945 mV

4, 104290

The higher the typical resistance values of circuits in a fault situation, the greater the difference between curves a and b, and the easier it is to find a faulty circuit.

Fig. 3a shows a measuring circuit 5 corresponding to the method according to the invention for four test circuits C1 ... C4. On a circuit-by-circuit basis, the daisy-chain connection is assumed to have a resistor of about 20 ohms. The diodes D connected to the circuits and the measured simulated voltages according to Fig. 3b with a current of 10 mA at the minimum resistance of the faulty circuit for reliable identification (108 ohms, estimated from the diagram) can be summarized in the table as follows:

Boundary D = IN5626 U / 108 ohm

10 1 1 pieces 1.02 V

2 2 pieces 1.35V

3 3 pcs 1.55V

4 7 pieces 1.65V

It is noted that seven diodes are connected across circuit C4. The reason for this figure is to take into account different fault combinations, since the simulation curves a ... d of Fig. 3b are obtained by assuming that only one circuit C1 ... C4 at a time is broken. If the two circuits over which one of the three diodes is connected fail, the voltage value becomes too close to the single defective circuit over which the four diodes are connected to identify the circuits. The corresponding pair is two and three diode circuits 20 and one five diode circuit. Thus, in the example of Figure 3a, the next reasonable number of diodes (six diodes could in principle be used) after the three diodes is seven to provide a sufficient voltage margin between the third and fourth circuits. Fault combinations require larger changes in resistance values than the aforementioned 108 ohms to reliably identify faulty circuits. For example, if all circuits except circuit C4 are broken, the sum of the voltages of circuits C1, C2 and C3 25 with a minimum resistance of 10 ohms in Fig. 3b will be of the order of 3V, the sum increasing quite steeply with 6V. The 3V voltage in Fig. 3b corresponds to a curve d of about 300 ohms, thus being slightly inclined towards 3.5V. Thus, if the resistance change in the circuit C4 is at least 300 ohms, the dissipation of the circuit C4 due to the difference in the voltage curves can be further distinguished from the worst-case combination C1-C3 of the other circuits.

It will be apparent to one skilled in the art that various embodiments of the invention are not limited to the examples set forth above, but that they may vary within the scope of the following claims. Thus, in addition to the above thermal cycle test, the method of the invention is particularly applicable to all component reliability tests where failure causes a break or a radical change in resistance.

Claims (3)

104290 A method for testing the reliability of integrated circuits, wherein in each test circuit (C1-C4) the junctions (1) are interconnected as desired (2a, 2b) to provide a resistor suitable for testing purposes in the circuit, and the test circuits are connected to to feed through the test circuits simultaneously, the feeling 11 u that a different number of diodes (D) downstream of the measuring current are connected across each test circuit (C1-C4) such that when the voltage across the circuit changes due to circuit breakage or other failure a circuit break or an increase in internal resistor causes the diodes connected over it to enter a conductive state, the threshold voltages of the diodes forming a characteristic voltage across the broken circuit which allows the faulted circuit to be detected by voltage measurement across all circuits. Method according to Claim 1, characterized in that in each of the 15 circuits to be tested, the joints (1) are connected together in a so-called. daisy-chain method, in which adjacent joints are connected alternately within the circuit (2b) and alternately from the substrate side (2a). Method according to claim 1 or 2, characterized in that during the test, a voltage is measured with each of the circuits under test (C1-C4) over a resistor connected to the circuit to determine the condition of the circuit. , 104290