JP4387706B2 - Grinding apparatus and grinding method - Google Patents

Description

[0001]
[Technical field to which the invention belongs]
The present invention relates to a grinding apparatus and a grinding method that can be used for surface grinding of, for example, a semiconductor wafer and a hard disk, and more particularly to improvement of a technique for controlling the surface position or thickness of a workpiece to a target value.
[0002]
[Prior art]
Conventionally, in equipment that grinds semiconductor wafers using a grindstone, the surface position (or thickness) of the workpiece being processed is constantly monitored using a contact sensor, and the measured value of the sensor is the target value. Generally, a method of changing or stopping the processing speed (grindstone feed speed) is generally performed.
[0003]
Moreover, the thing of patent document 1 is known as a system which correct | amends the cutting amount of a grindstone in consideration of the abrasion amount of a grindstone. According to the method described in Patent Document 1, a grindstone is fed to a previous machining completion position with a preset machining pattern, and after the grindstone has reached that position, an additional cut is made by a preset amount of wear of the grindstone. To do.
[0004]
Further, as a method of correcting the cutting amount of the grindstone in consideration of the wear amount of the grindstone and the thermal deformation of the feed shaft of the grindstone, the method described in Patent Document 2 is known. When the workpiece is machined on the side surface of the disk-shaped grinding wheel while changing the relative position between the rotating disk-shaped grinding wheel and the workpiece according to preset control data, The amount of wear of the grinding wheel and the amount of displacement of the grinding wheel in the rotational axis direction are detected in parallel with the processing by the sensor, and the workpiece is applied to the workpiece according to the detected amount of grinding wheel wear and the amount of displacement of the grinding wheel in the rotational axis direction. Correct the depth of cut.
[0005]
[Patent Document 1]
Japanese Patent No. 2934027
[Patent Document 2]
JP-A-8-243905
[0006]
[Problems to be solved by the invention]
When a contact-type sensor is used during processing, the contact will rub the surface of the rotating workpiece, and the workpiece is damaged. In addition, chips and peeling abrasive grains of the workpiece are mixed with the grinding fluid and flow toward the sensor contacts on the workpiece surface. These particles are caught in the contact portion of the sensor contact, and the sensor contact is damaged by these particles for a long period of time. And the damage | wound attached | subjected to the sensor contact will give a damage | wound to a workpiece. In particular, when the grindstone contains diamond powder, the damage of the sensor contact is fast.
[0007]
In the method described in Patent Document 1, after the grindstone is sent to the previous processing completion position, the grindstone is further fed at the grinding speed by the amount of wear of the grindstone. According to this method, the machining pattern (grinding wheel feed rate control pattern) changes in accordance with the amount of wear of the grinding wheel. Therefore, when machining at multiple grinding speeds, there is a problem that the machining conditions change every time. Further, according to this method, only the distance of the grinding portion in the processing pattern is extended according to the wear amount of the grindstone, so that the processing time increases (decrease in production efficiency) occurs.
[0008]
In addition, according to the method described in Patent Document 2, a sensor for detecting wear of the grindstone is required in addition to the sensor for measuring the surface position of the workpiece, so that the installation cost of the sensor is excessive. In this method, it is assumed that the positional relationship between the sensor and the machine is always maintained. However, in an environment where the grinding fluid is applied to the sensor, it is difficult to maintain this assumption, leading to an increase in cost.
[0009]
Accordingly, an object of the present invention is to allow the surface position (or thickness) of the workpiece to be controlled to a target value without constantly monitoring the surface position (or thickness) of the workpiece during processing, It is to reduce the maintenance cycle and cost by reducing the damage to the workpiece, improving the yield of the workpiece, and extending the life of the sensor contact.
[0010]
Another object of the present invention is to use a non-contact type sensor that is easily affected by polishing liquid or the like and is generally inferior to a contact type sensor as a sensor for measuring the surface position (or thickness) of the workpiece. There is in doing so.
[0011]
Another object of the present invention is to improve production efficiency.
[0012]
[Means for Solving the Problems]
The apparatus (10) for grinding the surface of the workpiece (12) with the grindstone (13) while feeding the grindstone (13) according to one aspect of the present invention is the surface position or thickness of the workpiece (12). A sensor (17) that generates a sensor measurement value (S) and a feed position detection means (16) that detects a feed position of the grindstone (13) and generates a feed position value (X); And a control means (16) for controlling the feed speed (Vp) of the grindstone (13), and the control means (16) performs the processing of the previous processing times in each processing time when a plurality of processing times are continuous. Means (105) for obtaining a correction amount (Δ (N)) related to the grinding wheel wear amount generated up to the previous machining round based on the sensor measurement value (S) and the feed position value (X) acquired at the completion. ) And acquired in real time during processing at each processing time Of the workpiece (12) being processed based on the feed position value (X (t)) and the correction amount (Δ (N)) related to the grinding wheel wear amount generated up to the previous processing round. Means (111) for obtaining an estimated value (s (t)) of the surface position or thickness, and in each processing round, based on the estimated value (s (t)) of the surface position or thickness, Grinding wheel (13) feed speed (Vp) And when processing is complete (111) which controls.
[0013]
According to this grinding processing apparatus, it is not necessary to constantly monitor the surface position or thickness of the workpiece with a sensor during processing. Therefore, even if a contact-type sensor is used, the sensor can be kept away from the workpiece during processing. Further, a non-contact type sensor that is not suitable for measurement during processing using a grinding fluid can also be used.
[0014]
In a preferred embodiment, the control device (16) further includes means for acquiring an estimated value (m) of the grinding wheel wear amount generated during the current machining, and the estimated value (m) of the grinding wheel wear amount is obtained. In addition, the feed speed (Vp) of the grindstone (13) is controlled in consideration.
[0015]
According to this configuration, it is possible to perform control with higher accuracy in consideration of the grinding wheel wear amount generated during the current machining.
[0016]
In a preferred embodiment, the control device (16) includes means for obtaining a correction amount (d (N-1) or k · d (N-1)) for suppressing a variation correction of the machining amount. In addition, the feed rate (Vp) of the grindstone (13) is controlled in consideration of the correction amount (d (N-1) or k · d (N-1)).
[0017]
According to this configuration, fluctuations (variations) in each processing time are suppressed, and a stable processing result can be obtained.
[0018]
In a preferred embodiment, the control unit (16) further includes proximity detection means for detecting that the grindstone (13) is close to the workpiece (12) while the grindstone (13) is being fed. ) Controls the position (Xf) at which the machining is completed based on the detection result of the proximity detection means.
[0019]
According to this configuration, the processing completion position can be controlled based on the position where the grindstone is close to the workpiece (12).
[0020]
In a preferred embodiment, the control unit (16) further includes proximity detection means for detecting that the grindstone (13) is close to the workpiece (12) while the grindstone (13) is being fed. ) Controls the timing at which the feed speed (Vp) is reduced to a safe speed for avoiding impact contact based on the detection result of the proximity detection means.
[0021]
According to this configuration, it is possible to reduce the idle running distance in which the grindstone is approached to the workpiece at a low safe speed, so the idle running time is shortened and the production efficiency is improved.
[0022]
DETAILED DESCRIPTION OF THE INVENTION
The present invention can be applied to various types of grinding processing apparatuses. In the following, for the purpose of explaining the present invention, for example, the present invention is applied to a surface grinding apparatus used for grinding a surface of a semiconductor wafer. An embodiment will be described as an example.
[0023]
[First Embodiment]
FIG. 1 shows a schematic configuration of a main part of a surface grinding apparatus according to an embodiment of the present invention. FIG. 2 is a diagram for explaining a basic method for feeding a grindstone in this embodiment.
[0024]
As shown in FIG. 1, this surface grinding apparatus 10 is disposed so as to face a table (chuck) 11 for holding and rotating a semiconductor wafer 12 as a workpiece on the surface by suction, for example. The grinding wheel feeding mechanism 14 for feeding the grinding stone 13 for grinding so as to approach or move away from the table 11, and the sensor 17 for measuring the thickness or surface position of the semiconductor wafer 12 mounted on the table 11. With. The type of the sensor 17 may be a contact type or a non-contact type, but in this embodiment, it is a contact type, and the contacts 17A and 17B can be raised and lowered. For example, as shown in FIG. While the thickness or surface position of the wafer 12 is being measured, the contacts 17A and 17B are lowered to contact the table (chuck) 11 and the surface of the semiconductor wafer 12, while the measurement is performed as shown in FIG. While not being performed, the contacts 17A and 17B are raised so as to be separated from the surface of the table (chuck) 11 and the semiconductor wafer 12. The grindstone feed mechanism 14 reciprocates the grindstone feed shaft 15 with the grindstone 13 attached to the tip and the grindstone feed shaft 15 in the axial direction (left and right in the figure), and the grindstone feed shaft 15 is rotated around its central axis. And a feed shaft driving device 16 for controlling the position and feed speed of the grindstone feed shaft 15. The reciprocating direction of the grindstone feed shaft 15 (feed direction of the grindstone 13) is perpendicular to the surface of the table 11 (the surface on which the semiconductor wafer 12 is fixed). The front surface of the grindstone 13 is parallel to the surface of the table 11.
[0025]
The sensor 17 measures the thickness or surface position of the semiconductor wafer 12 and outputs the measurement result as a numerical value (hereinafter referred to as “sensor measurement value”) S (t). The feed shaft driving device 16 has a feed position detector (not shown). The feed position detector measures the displacement amount (feed amount) of the grindstone feed shaft 15 (grinding stone 13), and displays the measurement result. Based on this, the position of the grindstone feed shaft 15 (grindstone 13) (hereinafter referred to as “feed position”) is output as a numerical value X (t) (hereinafter referred to as “feed position value”). As a configuration for sending the grindstone 13, instead of a configuration in which the position of the table 11 is fixed and the grindstone 13 is moved as illustrated, a configuration in which the position of the grindstone 13 is fixed and the table 11 is moved, or A configuration in which both the grindstone 13 and the table 11 are moved can also be employed. In any configuration, the feed position detector outputs a feed position value X (t) representing the feed position of the grindstone 13 by measuring a relative displacement amount (feed amount) of the grindstone 13 with respect to the table 11. To do. The feed shaft driving device 16 uses the feed position value X (t) from the feed position detector and the sensor measurement value S (t) from the sensor 17 to feed the grinding wheel feed shaft 15 (grindstone 13). Control Vp.
[0026]
The general operation during the machining operation is as follows. The semiconductor wafer 12 is fixed on the table 11, and the contacts 17A and 17B of the sensor 17 are lowered as shown in FIG. 2A, and the thickness or surface position of the semiconductor wafer 12 on the table 11 before processing is measured. The This measured value is used to correct a processing pattern used for processing the semiconductor wafer 12 (details will be described later). Subsequently, as shown in FIG. 2B, the sensor contacts 17A and 17B are raised, and then the table 11 is rotated to rotate the semiconductor wafer 12, and the feed shaft driving device 16 is moved to the grindstone feed shaft. While rotating 15, the grindstone feed shaft 15 is moved forward (left direction in the figure) from the most retracted position (that is, the grindstone 13 is fed forward). As a result, the grindstone 13 approaches the semiconductor wafer 12 on the table 11 while rotating, and eventually, the front surface of the grindstone 13 contacts the surface of the semiconductor wafer 12 to start grinding. The grindstone 13 is further fed to grind the surface of the semiconductor wafer 12. Until the grinding process is completed, the contacts 17A and 17B of the sensor 17 remain raised as shown in FIG. 2B. During the grinding process, the contacts 17A and 17B are based on the feed position value X (t). An estimated value s (t) of the sensor measurement value S (t) is calculated. When the estimated value s (t) of the sensor measurement value reaches a predetermined processing end position, the forward feeding of the grindstone 13 is stopped and the grinding processing is completed. After the grinding process is completed, after the rotation of the semiconductor wafer 12 is stopped, the sensor contacts 17A and 17B are lowered again as shown in FIG. 2A, and the thickness or surface position of the semiconductor wafer 12 after the completion of the process is determined. It is measured. This measured value is used to correct a processing pattern used in the next processing round (that is, processing of the next semiconductor wafer 12) (details will be described later). The above operation is repeated every processing time.
[0027]
While the grindstone 13 (grindstone feed shaft 15) is being fed as described above, the feed shaft driving device 16 feeds the grindstone 13 (grindstone feed shaft 15) according to a preset processing pattern PP as shown in FIG. Control the speed Vp. This machining pattern PP is data instructing how to control the feed speed (machining speed) Vp (for example, which target position should be controlled to which target speed). Target speeds and a plurality of target positions corresponding to these target speeds (this indicates a position where the target speed should be switched, and is hereinafter referred to as “feed speed switching position”). While the feed shaft driving device 16 controls the feed speed Vp according to the machining pattern PP, the feed position value X (t) and the machining pattern before the grindstone 13 contacts the semiconductor wafer 12 (that is, before the grinding process is started). By comparing the PP feed speed switching position, and after the grindstone 13 contacts the semiconductor wafer 12 (that is, after the grinding process is started), the estimated value s (t) of the sensor measurement value S (t) and the processing By comparing with the feed speed switching position of the pattern PP, it is determined whether or not the grindstone 13 has reached the feed speed switching position, and when it reaches, the target value of the feed speed Vp is set to the next target according to the machining pattern PP. Switch to the value. In this embodiment, the origin of the feed rate switching position of the machining pattern PP is the surface position of the table 11 from the viewpoint of ease of setting, and similarly, the origin of the sensor measurement value S (t) (S = 0) and the origin (X = 0) of the feed position value X (t) are initially set to the surface position of the table 11.
[0028]
For example, according to the specific example of the machining pattern PP shown in FIG. 3, the target speed of the feed speed Vp is initially the fastest fast feed speed Vpa and then switched to the slower buffer feed speed Vpb. Thereafter, when the feed position value X (t) reaches a predetermined idle running start position Xpa, the final approach of the grindstone 13 to the semiconductor wafer 12, that is, idle running is started. In general, it is desirable to feed the grindstone 13 at a speed as low as the grinding speed (up to several times the machining speed at most) in order to reduce the contact impact on the semiconductor wafer 12 during idle running. For example, in the specific example shown in FIG. 3, during idling, the same roughing speed Vpc as during rough grinding is set as the target speed. When the grindstone 13 is sent by a certain distance Lfr (hereinafter referred to as “idle running distance”) from the idling start position Xpa, it contacts the semiconductor wafer 12. Grinding is started from here, but for a while after the start of machining, rough grinding is performed while maintaining the feed speed Vp at the roughing speed Vpc. Thereafter, the feed speed Vp is switched to a lower finishing speed Vpd, and the final finish grinding is performed. Thereafter, when the estimated value s (t) of the sensor measurement value S (t) reaches the machining end position Xps, the grinding process is completed.
[0029]
As described above, the origin (S = 0) of the sensor measurement value S (t) and the origin (X = 0) of the feed position value X (t) are initially set to the surface position of the table 11. However, after the initial setting, an offset (hereinafter referred to as “sensor origin offset”) Sofs occurs between the origin (S = 0) of the sensor measurement value S (t) and the surface position of the table 11, or the feed position value X An offset (hereinafter referred to as “feed axis origin offset”) Mofs also occurs between the origin (X = 0) of (t) and the surface position of the table 11. The causes of these offsets Sofs and Mofs are thermal deformation of the machine due to temperature changes and elastic deformation of the machine due to processing load. These offsets Sofs and Mofs change over time until the thermal balance is completely achieved. . Conventionally, in order to avoid this change, warm-up operation of about several hours may be performed. Further, the wear amount of the grindstone 13 increases with the processing time. The changes in the amount of wear of the offset Sofs, Mofs and the grindstone 13 are so large that they cannot be ignored compared to the processing accuracy.
[0030]
In order to solve the above-described problem, the feed shaft driving device 16 performs processing patterns according to the offset Sofs, Mofs and the wear amount of the grindstone 13 at each processing time (opportunity to process one semiconductor wafer 12). It has a function to correct PP to the optimum one. Further, even if the feed shaft driving device 16 raises the contacts 17A and 17B of the sensor 17 away from the semiconductor wafer 12 during the processing operation, the thickness or surface position of the semiconductor wafer 12 becomes a target value. It has a function of appropriately controlling the feeding operation of the grindstone 13. Hereinafter, machining control by the feed shaft driving device 16 incorporating these functions will be described.
[0031]
FIG. 4 is a diagram showing a processing control procedure by the feed shaft driving device 16. FIG. 5 is a diagram for explaining how to correct the processing pattern PP.
[0032]
1. Origin setting of sensor measurement value S (t) and feed position value X (t) (step 101 in FIG. 4)
At a certain reference time point, the contacts 17A and 17B of the sensor 17 are lowered and brought into contact with the surface of the table 11, and the front surface of the grindstone 13 is brought into contact with the surface of the table 11. In this state, the sensor measurement value S (t) and The feed position value X (t) is acquired at the same time. These acquired value pairs are stored as origin value pairs (S (0), X (0)) representing the origin of the sensor measurement value S (t) and the feed position value X (t) (that is, the table Initially set the surface position as the origin).
[0033]
A place other than the surface of the table 11 can be set as the origin for obtaining the origin value pair. That is, if it is a place that can be used as a reference point for measuring the surface position or thickness of the semiconductor wafer 12 on the table 11, for example, if it is a place where the relative positional relationship with the table 11 is known, the position Can be set as the origin. For example, a jig having a known thickness may be set on the surface of the table 11 and the position of the surface of the jig may be set as the origin. In this case, the grindstone 13 or the like does not contact the surface of the table 11. Further, it is possible to avoid the surface of the table 11 from being damaged due to contact.
[0034]
When the above origin setting is completed, the contacts 17A and 17B of the sensor 17 are raised and separated from the table 11, and the grindstone 13 is retracted to a predetermined feed start position far from the table 11.
[0035]
2. First processing (steps 102 and 103 in FIG. 4)
A first semiconductor wafer 12 is fixed to the table 11. One sensor 17A of the sensor 17 is brought into contact with the surface of the semiconductor wafer 12, and the other contact 17B is brought into contact with the surface of the table 11, and the sensor 17 is set in a state in which the thickness or surface position of the semiconductor wafer 12 is measured. (After that, the sensor 17 is maintained in this state until the first machining is completed). Then, “1” is set in the processing time N, and a standard processing pattern PP_std (see FIG. 5) prepared in advance is set as a processing pattern used in the first processing (step 102). Then, the first machining operation is started (step 103). Therefore, in the first machining operation, the grindstone 13 is sent according to the standard machining pattern PP_std. Here, the standard processing pattern PP_std can be set by a conventional method, for example, by a user. According to the conventional method, the free running start position Xpa_std in this standard processing pattern PP_std is due to the maximum value of the thickness of the semiconductor wafer 12 (because the actual thickness varies around the nominal thickness), thermal deformation, elastic deformation, etc. It is set to a safe position (a position where the grindstone 13 is not likely to contact the semiconductor wafer 12) in consideration of the maximum elongation of the grindstone feed shaft 15 or the like. In the first processing, until the grindstone 13 comes into contact with the semiconductor wafer 12, the grindstone 13 comes into contact with the semiconductor wafer 12 by comparing the feed position value X (t) with the speed switching position of the machining pattern PP_std. After (during machining), it is determined whether the grindstone 13 has reached each target position by comparing the sensor measurement value S (t) with the speed switching position of the machining pattern PP_std. Based on the determination result, The feed speed Vp of the grindstone 13 is controlled to the target speed corresponding to the machining pattern PP_std.
[0036]
3. Acquisition of first machining completion point (step 104 in FIG. 4)
When the first grinding process is substantially completed, the sensor 17 measures the surface position of the semiconductor wafer 12 and the sensor measurement value in a state where the front surface of the grindstone 13 is in contact with the surface of the semiconductor wafer 12. S (t) and feed position value X (t) are acquired at the same time, and these acquired value pairs are used as the first processing completion point value pair (the surface position of the semiconductor wafer 12 when the first grinding processing is completed) S (1), X (1)). The above-mentioned “when the grinding process is substantially completed” may be the time when the grinding process is finished 100% or may be the time when the grinding process is finished close to 100%. For example, in order to remove the scratches on the surface of the semiconductor wafer 12 caused by the sensor contact 17A, the sensor contact 17A can be lifted slightly before the completion of the processing, and then a slight finishing process can be additionally performed. In such a case, the first machining completion point value pair (S (1), X (1)) is acquired immediately before raising the sensor contact 17A (for example, when machining is finished by about 90%). Will do. Alternatively, in order to make the measurement conditions the same as the second and subsequent times, the feed position value X (t) is acquired while the grindstone 13 remains in contact with the surface of the semiconductor wafer 12 when the processing is completed, and the sensor contact The children 17A and 17B are again lowered and brought into contact with the surface of the semiconductor wafer 12 and the table 11, and a sensor measurement value S (t) representing the thickness or surface position of the semiconductor wafer 12 at the completion of the processing is obtained and processed for the first time. A completion point value pair (S (1), X (1)) may be used.
[0037]
4). Shift of the standard machining pattern PP_std (steps 105 and 108 in FIG. 4) The correction amount Δ (1) is obtained by the following equation 1 using the data acquired at the first machining operation (step 105).
[0038]
Δ (1) = − ((X (1) −X (0)) − (S (1) −S (0)))
=-((X (1) -S (1))-(X (0) -S (0))) (Formula 1)
The meaning of Formula 1 is as follows. The term (S (1) -S (0)) represents the surface position of the semiconductor wafer 12 when the first grinding process is completed with the surface position of the table 11 as the origin. Further, the term (X (1) −X (0)) represents the surface position of the semiconductor wafer 12 when the first grinding process is completed with the surface position of the table 11 as the origin if the grindstone 13 is not worn. Actually, however, because the grinding wheel 13 is worn, the position closer to the table 11 than the surface position of the semiconductor wafer 12 when the first grinding process is completed by the amount of wear W (1) of the grinding wheel 13. Represents. Furthermore, if there is no thermal deformation or elastic deformation of the machine (mainly the grindstone feed shaft 15), both X (0) and S (0) represent the surface position (origin) of the table 11, but actually Because of such mechanical deformation, the positions deviated from the surface position (origin) of the table 11 by the offset Xofs and Sofs due to the thermal deformation are shown. Therefore, if there is neither grinding wheel wear nor mechanical deformation, the term (X (1) -X (0)) and the term (S (1) -S (0)) in Equation 1 are equal and the correction amount Δ (1 ) Is zero, but in reality, there is grinding wheel wear and mechanical deformation, and the value of that amount appears in the correction amount Δ (1). That is, the correction amount Δ (1) is the total deformation amount (in other words, in the first machining) by combining the grinding wheel wear amount W (1) generated in the first grinding process and the error amount E (1) due to mechanical deformation. This represents a substantial shortening amount of the entire length of the grindstone feed shaft 15 including the grindstone 13 due to the generated grindstone wear and mechanical deformation.
[0039]
In order to optimize the machining pattern for the second machining operation in accordance with the substantial shortening amount of the grindstone feed shaft 15 in the first grinding process, that is, the correction amount Δ (1), FIG. As indicated by a one-dot chain line arrow, the entire standard machining pattern PP_std is shifted toward the table 11 by the correction amount Δ (1) (that is, the correction amount Δ (1) from each speed switching position in the standard machining pattern PP_std). (Step 108).
[0040]
5. Correction of idling start position Xpa (step 109 in FIG. 4)
The second semiconductor wafer 12 is fixed to the table 1, and before starting the second processing operation, the sensor contacts 17A and 17B are lowered to contact the surface of the table 11 and the semiconductor wafer 12, and the sensor 17 causes the semiconductor wafer to be contacted. 12 surface positions are measured. From the sensor measurement value Sa (2), the idling start position Xpa (2) for the second machining is obtained by the following equation 2.
[0041]
Xpa (2) = (Sa (2) −S (0)) + X (0) −Δ (1) + MA (Formula 2)
The meaning of Formula 2 is as follows. The first term (Sa (2) -S (0)) is the surface position before grinding of the second semiconductor wafer 12 with the surface position of the table 11 as the origin (that is, the second semiconductor wafer 12 Thickness before grinding). The origin X (0) of the feed position value X (t) is added to this value, and the correction amount Δ (1) (that is, the grindstone feed shaft 15 including the grindstone 13 due to grinding wheel wear, mechanical deformation, etc.). Is subtracted), the feed position value X (t) in a state where the front surface of the grindstone 13 is in contact with the surface position of the second semiconductor wafer 12 before grinding is obtained. . By adding an appropriate margin MA for avoiding shock contact between the semiconductor wafer 12 and the grindstone 13 to this value, the idling start position Xpa (2) for the second machining operation is determined. Here, the margin MA is a value corresponding to the idling distance Lfr, and may be a value of the order of correction accuracy in this embodiment, and can be set appropriately by the user or the like.
[0042]
Using the idling start position Xpa (2) obtained by Expression 2, a portion related to the idling start position Xpa in the machining pattern PP ′ (2) corrected in step 108 is corrected. That is, for example, the idling start position Xpa in the corrected machining pattern PP ′ (2) is replaced with the idling start position Xpa (2) obtained by the equation 2, and accordingly, the buffer feed speed immediately before that is changed. The speed switching position that determines the timing for starting to reduce the speed from Vpb is also corrected (dotted line arrow in FIG. 5). As a result, not only grinding wheel wear and mechanical deformation, but also an error D (2) (reflected in the first term (Sa (2) −S (0)) due to variations in the thickness of the semiconductor wafer 12 is taken into consideration. And the free running start position Xpa (2) is optimized. The above correction result is set as the machining pattern PP (2) for the second machining operation.
[0043]
Note that the second idling start position Xpa (2) may be obtained by the following formula 3 instead of the above formula 2.
[0044]
Xpa (2) = X (0) −Δ (1) + MA + THmax (Formula 3)
Here, the value THmax is a value obtained by adding the maximum value of the change amount of the sensor offset Sofs to the maximum thickness of the semiconductor wafer 12 predicted from the variation in the thickness of the semiconductor wafer 12.
[0045]
6). Second machining (steps 110 and 111 in FIG. 4)
2 is set in the machining time N (step 110). Then, the sensor contacts 17A and 17B are raised and separated from the surface of the table 11 and the semiconductor wafer 12, and then the table 11 and the grindstone 13 are rotated, and the feeding of the grindstone 13 is started, and the second processing described above. A second machining operation is performed using the machining pattern PP (2) for operation (step 111).
[0046]
In the second machining operation, until the grindstone 13 contacts the semiconductor wafer 12, the grindstone feed is compared with the feed position value X (t) and the speed switching position of the machining pattern PP (2) as in the first time. Take control. However, after the grindstone 13 contacts the semiconductor wafer 12, the grindstone feed control cannot be performed while comparing the sensor measurement value S (t) and the speed switching position of the machining pattern PP (2) as in the first time. . This is because the sensor 17 is separated from the semiconductor wafer 12 and the sensor measurement value S (t) cannot be obtained. Therefore, in the second machining operation, the estimated value s (t) of the sensor measurement value S (t) is obtained using the feed position value X (t) sequentially acquired in real time while the grindstone 13 is being fed. It is obtained in real time by sequential calculation according to the following equation, and the grindstone feed control is performed while comparing the estimated value s (t) with the speed switching position of the machining pattern PP (2).
[0047]
s (t) = (X (t) + Δ (1) −X (0)) + S (0) + m + d (1) (Formula 4)
The meaning of Equation 4 is as follows. The first term (X (t) + Δ (1) −X (0)) is the surface position during grinding of the second semiconductor wafer 12 with the surface position of the table 11 as the origin (that is, the second sheet The thickness of the semiconductor wafer 12 during grinding) is expressed using the feed position value X (t). By adding the origin S (0) of the sensor measurement value S (t) to this value, this value is converted into a value represented by the sensor measurement value S (t). Since this value includes the correction amount Δ (1), the amount of grinding wheel wear and the amount of mechanical deformation generated until the previous machining is completed are incorporated. However, this value does not yet take into account the amount of grinding wheel wear that will occur during the current machining. Therefore, an estimated value m of the grinding wheel wear amount that will occur during the current machining is added to this value, so that the surface position or thickness of the semiconductor wafer 12 is actually measured by the sensor 17 during the current grinding. An estimated value of the sensor measurement value S (t), which would be obtained if measured, is obtained. Furthermore, a correction value d () for suppressing fluctuations in the machining amount at each machining time (this is caused by variations in the grinding wheel wear amount and other conditions at each machining time) with respect to the sensor measurement value estimation value. 1) is added.
[0048]
Here, the grinding wheel wear amount that occurs during the current machining described above is actually a value that gradually increases as the machining progresses. Therefore, as the estimated value m of the grinding wheel wear amount, for example, the grinding wheel wear amount ΔW (N for one grinding process (that is, all generated from the start to the completion of one grinding process). ) Or half of the estimated value (that is, the estimated amount of wear at an intermediate point between the start and completion of one machining), or grinding wheel wear per grinding process The estimated value of the amount ΔW (N) can be a value that is proportionally distributed according to the feed distance from the start to the end of machining. Further, the grindstone wear amount estimated value m may be a fixed value prepared in advance, or may be a value calculated from values acquired at the time of previous machining. In addition, the grinding wheel wear amount estimated value m is not included in the sensor measurement value estimated value s (t). Instead, the target position during machining of the machining pattern (such as the machining end position or the position where the machining speed is switched) is worn. You may correct by the quantity estimated value m.
[0049]
The correction value d (N-1) for suppressing the machining amount fluctuation is, for example, that the previous machining amount (Nth) is larger than the target value of the machining thickness or depth. If it is too much, the machining amount is slightly reduced this time (Nth time). Conversely, if the amount of machining in the previous time (N-1th time) is too small, the machining amount is somewhat increased this time (Nth time). As described above, the value according to the deviation of the past machining amount is fed back to the future machining control, and used to reduce the deviation of the future machining amount. The correction value d (N-1) can be obtained by calculation based on the past machining amount such as the previous time or the previous time. For example, when the processing purpose is to keep the processed thickness of the semiconductor wafer 12 constant, the correction value d (N) is expressed as the difference between the previous and previous processed thicknesses as follows:
d (N-1) = S (N-1) -S (N-2) (Formula 5)
In addition, in the case of removing a certain depth from the original surface of the semiconductor wafer 12 (making the processing amount constant), the correction value d (N) is the processing amount of the previous time and the previous processing amount. The difference is
d (N-1) = (Sa (N-1) -S (N-1))-(Sa (N-2) -S (N-2)) (Formula 6)
Can be obtained. Here, Sa (N-1) and Sa (N-2) are the thicknesses of the semiconductor wafer 12 measured by the sensor 17 before the start of the previous (N-1th) and previous (N-2th) processing. Yes, S (N-1) and S (N-2) are the thicknesses of the semiconductor wafer 12 measured by the sensor 17 after completion of the previous (N-1th) and previous (N-2th) processing. In addition, this fluctuation suppression correction value d (N-1) is not included in the sensor measurement value estimation value s (t), similarly to the above-described grinding wheel wear amount estimation value m. May be corrected by the correction value d (N-1).
[0050]
However, the fluctuation suppression correction value d (N−1) obtained using the above-described expression 5 or 6 is used at the time of the third (N = 3) and subsequent processing, and the second (N = 2). The correction value d (1) used at the time of machining is zero, or S (1)-(target thickness) or (Sa (1) -S (1))-(target depth).
[0051]
7. Acquisition of second machining completion point (steps 112 and 104 in FIG. 4)
When the sensor measurement value estimated value s (t) obtained by the above equation 4 reaches the second machining completion position Xps (2) of the machining pattern PP (2), the second machining is completed. When the processing is completed, the feed position value X (t) is acquired while the grindstone 13 is still in contact with the surface of the semiconductor wafer, and the sensor contacts 17A and 17B are again lowered onto the surface of the semiconductor wafer and the table 11. The sensor measurement value S (t) representing the thickness or surface position of the semiconductor wafer at the completion of processing is obtained by contact. The acquired feed position value X (t) and sensor measurement value S (t) are stored as a second machining completion point value pair (S (2), X (2)).
[0052]
In the measurement by the sensor 17 at the completion of the processing, the rotation of the table 11 or the grindstone 13 is stopped, and in some cases, the grindstone 13 is further retracted to a position that does not affect the measurement, so that the sensor contact 17A, After eliminating the factor that 17B is damaged by abrasive grains or the like, the sensor contacts 17A and 17B can be lowered to perform measurement.
[0053]
8). Determination of the processing pattern after the third time (steps 105 to 109 in FIG. 4)
Thereafter, the processing pattern PP (N + 1) for the next time (N + 1) is optimized by performing the above-described steps 105 to 109 every processing time N (N = 2, 3, 4,...). That is, the machining completion point value pair ((S (N), X (N)) acquired at the completion of the grinding process (Nth) and the thickness of the semiconductor wafer 12 before the next (N + 1) grinding process. Using the sensor measurement value Sa (N + 1) obtained by measuring the height with the sensor 17, the correction amount Δ (N) for the next time (N + 1) and the idle running start position Xpa (N + 1) are expressed by Equation 7 and The next processing pattern PP (N + 1) is determined by calculating the equation (8) and using these values to modify the standard processing pattern PP_std by the method described above.
[0054]
Δ (N) = − ((X (N) −S (N)) − (X (0) −S (0))) (Expression 7)
Xpa (N + 1) = (Sa (N + 1) −S (0)) + X (0) −Δ (N) + MA (Expression 8)
[0055]
When the heat balance is completely achieved and there is no time change in the machine dimensions due to thermal deformation, the difference in the correction amount (Δ (N + 1) −Δ (N)) for each process is the grindstone per process. The amount of wear is ΔW (N). Therefore, while there is almost no temperature change and the heat balance is substantially complete, in step 108, instead of the above-described method of shifting the standard processing pattern PP_std by the correction amount Δ (N), one processing is performed. A method of shifting the previous processing pattern PP (N) by the grinding wheel wear amount ΔW (N) per rotation can also be adopted. In this case, the grinding wheel wear amount ΔW (N) per processing time is calculated by comparing the processing completion point value pairs (X (N), S (N)) and (X (N−1)) of the Nth time and the N−1th time. Since it can be calculated only from S (N-1)), it can be obtained as a reliable value that is not affected by elastic deformation or the like.
[0056]
9. Processing after the third time (steps 110 to 112, 104 in FIG. 4)
Thereafter, the above-described steps 110 to 112 and 104 are performed at every processing times N (N = 3, 4,...). In particular, in step 110, the processing is performed while feeding the grindstone 13 with the sensor 17 being separated from the table 11 and the semiconductor wafer 12. During machining, based on the feed position value X (t),
s (t) = (X (t) + Δ (N−1) −X (0)) + S (0) + m + d (N−1) (Equation 9)
The sensor measurement value estimated value s (t) is obtained in real time by sequential calculation according to, and the grindstone feed control is performed while comparing the estimated value s (t) with the speed switching position of the machining pattern PP (N). . At this time, as already described, the correction value d (N) for suppressing the variation in the machining amount is expressed by the following equation:
d (N-1) = S (N-1) -S (N-2) (Formula 5)
Or
d (N-1) = (Sa (N-1) -S (N-1))-(Sa (N-2) -S (N-2)) (Formula 6)
You can use what you asked for.
[0057]
9-2. Gain adjustment for correction value d (N) to suppress machining amount fluctuation
The correction value d (N) calculated by Equation 5 or Equation 6 described above represents the machining completion position or the machining amount variation amount between the previous time and the previous time. In the above-described equation 9, the correction value d (N) is used as it is. As a result, the processing completion position or the amount of change in the processing amount decreases each time the processing times are followed, and eventually converges to almost zero. On the other hand, the accuracy of the pretreatment or preprocessing of the semiconductor wafer 12 or the grinding wheel Depending on the processing conditions such as the material No. 13 and the grindstone feed speed Vp, the fluctuation amount may not be small as shown in the dotted line graph of FIG. In the latter case, the adjustment value d (N) can be multiplied by an appropriate gain k (0 <k <1) in order to stably stabilize the fluctuation amount as shown in the solid line graph of FIG. That is, instead of the above equation 9, the following equation can be used.
[0058]
s (t) = (X (t) + Δ (N−1) −X (0)) + S (0) + m + k · d (N−1) (Equation 10)
[0059]
10. Correction of correction amount Δ (N) when there is a long interval (steps 106 and 107 in FIG. 4)
Continuous machining that repeats machining operations at short intervals is interrupted, and the interval from the completion of the previous (N times) grinding operation to the start of the next (N + 1) machining operation occurs during that time. If the deformation amount is long enough that it cannot be ignored (for example, when a predetermined threshold time T is exceeded), the previous machining completion point value pair (X (N), S (N ), The correction amount Δ (N) is corrected by Equation 6 by an adjustment amount δ set in advance in consideration of the amount of thermal deformation during the interval.
[0060]
Δ (N) = Δ (N) −δ (Formula 6)
[0061]
With this modification, a more optimized machining pattern is obtained according to the thermal deformation during the interval.
[0062]
The adjustment amount δ may be variably set according to the actual interval length based on a predetermined relationship between the interval length and the thermal deformation amount, instead of using a fixed value.
[0063]
In order to clarify the advantage of the correction method of the processing pattern PP (N) in the processing control described above, some supplementary explanation regarding the above calculation formula will be described below.
[0064]
When the true feed position and true sensor measurement position with the surface of the table 11 as the origin are represented by TX (t) and TS (t), respectively, the true feed position TX (N) and true Sensor measurement position TS (N) is
TX (N) = X (N) + Mofs (N) (Formula 11)
TS (N) = S (N) + Sofs (N) (Formula 12)
It becomes. If the grindstone 13 is not worn, the true feed position TX (N) at the completion of machining and the true sensor measurement position TS (N) are both the same position (the surface of the semiconductor wafer 12) regardless of the machining time N. Position),
TX (N) -TS (N) = 0 (Expression 13)
It becomes. However, in actuality, grinding wheel wear occurs, and the true feed position TX (N) advances toward the table 11 by the grinding wheel wear amount W (N) from the true sensor position TS (N).
TX (N) -TS (N) =-W (N) (Formula 14)
The difference will come out. In addition, since the offsets Mofs and Sofs change with time due to thermal deformation and elastic deformation of the machine, even if there is no change between the true feed position TX (N) and the true sensor position TS (N), The feed position X (N) and the sensor measurement value S (N) acquired in the above change with time as the offsets Mofs and Sofs change.
[0065]
On the other hand, at the reference timing N = 0, the grinding wheel wear amount W (0) is zero.
TX (0) -TS (0) = 0 (Formula 15)
Is established. Using equation 15 and rewriting equation 14,
W (N) =-((TX (N) -TS (N))-(TX (0) -TS (0))) (Equation 16)
When substituting Equations 11 and 12 into the right side of Equation 16,
W (N) = − (X (N) −S (N)) + (X (0) −S (0)) − (Mofs (N) −Mofs (0)) − (Sofs (N) −Sofs ( 0)) (Formula 17)
Is obtained. And when substituting equation 7 for the right side of equation 17,
Δ (N) = W (N) + (Mofs (N) −Mofs (0)) + (Sofs (N) −Sofs (0)) (Equation 18)
Is obtained.
[0066]
This equation 18 represents the meaning of the correction amount Δ (N) used in the control method of the present embodiment. That is, the correction amount Δ (N) includes not only the grindstone wear amount W (N) but also the temporal variation of the offset Mofs and Sofs due to thermal deformation and elastic deformation of the machine (Mofs (N) −Mofs (0 )), (Sofs (N) -Sofs (0)). In the control method of the present embodiment, correction is performed such that the processing pattern PP_std is shifted to the table 11 side by the correction amount Δ (N) every processing time. The specific meaning of this correction is as follows.
[0067]
First, the first component of the correction amount Δ (N) expressed by Equation 18 is the grindstone wear amount W (N). When the grindstone 13 is worn, the entire length of the grindstone feed shaft 15 including the grindstone 13 is shortened by the amount of the grindstone wear W (N). Therefore, in this control method, the machining pattern PP_std) is shifted to the table 11 side by the grinding wheel wear amount W (N) in order to compensate for this shortening.
[0068]
Next, the second component of the correction amount Δ (N) expressed by Equation 18 is the amount of change in the feed axis origin offset Mofs (Mofs (N) −Mofs (0)), which includes the grindstone 13. This includes thermal deformation and elastic deformation of the grindstone feed shaft 15, the table 11, and a base on which they are fixed. This amount of change in offset (Mofs (N) −Mofs (0)) can be regarded as substantially the expansion and contraction of the entire length of the grindstone feed shaft 15 including the grindstone 13 due to thermal deformation, elastic deformation, or the like. When the offset change amount (Mofs (N) −Mofs (0)) is a positive value, it means that the entire length of the grindstone feed shaft 15 including the grindstone 13 is substantially shortened by this value. . Therefore, in this control method, the machining pattern PP_std is shifted to the table 11 side by an offset change amount (Mofs (N) −Mofs (0)) in order to compensate for this shortening. On the contrary, when the offset change amount (Mofs (N) −Mofs (0)) is a negative value, it means that the entire length of the grindstone feed shaft 15 is substantially extended by that value. PP_std is shifted in the opposite direction to the table 11 by that value.
[0069]
Next, the third component of the correction amount Δ (N) expressed by Equation 18 is the change amount (Sofs (N) −Sofs (0)) of the sensor origin offset Sofs, which is caused by the sensor 17 or the table 11. And thermal deformation and elastic deformation of the base on which they are fixed. This offset change amount (Sofs (N) −Sofs (0)) is substantially the origin position deviation of the sensor 17. In general, various attempts have been made to prevent this origin position deviation, but there may be an origin position deviation that cannot be ignored. When this value is a positive value, it means that the origin position of the sensor 17 is shifted toward the grindstone 13 by that amount, and as a result, the surface position of the semiconductor wafer 12 measured by the sensor 17 is The actual surface position is shifted to the table 11 side by the value. Therefore, in this control method, the machining pattern PP_std is shifted to the table 11 side by an offset change amount (Sofs (N) −Sofs (0)) in order to compensate for this shift. On the contrary, when the offset change amount (Sofs (N) −Sofs (0)) is a negative value, the surface position of the semiconductor wafer 12 measured by the sensor 17 is more than the actual surface position. 11 is shifted in the opposite direction to the machining pattern PP_std, the machining pattern PP_std is shifted in the opposite direction to the table 11.
[0070]
In this way, according to this control method, not only the grinding wheel wear amount W (N), but also the dimensional change caused by thermal deformation and elastic deformation of the machine is taken into account to correct the machining pattern, so the thermal balance is perfect. Even if it is not removed, it is possible to enter the machining operation, and the warm-up operation time is shorter than that in the conventional control method that considers only the grinding wheel wear amount.
[0071]
When the thermal balance is achieved during the machining operation, for example, using Mofs (N) = Mofs (N−1), Sofs (N) = Sofs (N−1), The following equation 19 can be used to calculate the grinding wheel wear amount ΔW (N) per process. The next processing pattern PP (N + 1) can be optimized using this value ΔW (N).
[0072]
ΔW (N) = W (N) −W (N−1) = Δ (N) −Δ (N−1)
=-((X (N) -S (N)) + (X (N-1) -S (N-1)) (Equation 19)
[0073]
Further, according to the control method of the present embodiment, since the idle running start position Xpa is determined from the thickness of the semiconductor wafer 12 before processing, it is almost necessary to exclude the thermal deformation amount of the machine and the grinding wheel wear amount. By setting the minimum margin distance MA (about the correction accuracy for one time), the processing can be performed without damaging the semiconductor wafer 12. In other words, each idle running distance Lfr can be reduced to a necessary minimum, thereby reducing the overall work time. The control method of the present embodiment is particularly effective for improving throughput in processing that requires high processing accuracy and is inevitably performed under the conditions of low-speed processing and a large amount of grinding wheel wear (there are many precision processing). Can contribute.
[0074]
Further, according to the control method of the present embodiment, when the work stop time becomes long, the adjustment value δ is introduced, so that the processing is performed at a time when the grindstone feed shaft 15 and the like are considered to be thermally deformed. Even if the process is started, the semiconductor wafer 12 can be processed without fear of causing problems such as damage and quality deterioration.
[0075]
Also, according to the control method of the present embodiment, measurement by the sensor 17 is not performed during machining, and instead, the feed shaft position X (t), the grinding wheel wear amount and mechanical deformation that occurred until the previous machining are matched. The sensor measurement value S (t) is estimated in real time using the amount Δ (N−1) obtained and the estimated value m of the grinding wheel wear amount generated this time, and the estimated value s (t) is determined as the machining pattern. Since this is compared with PP (N), damage to the sensor contacts 17A and 17B can be reduced.
[0076]
Further, according to the control method of the present embodiment, a stable and constant machining result can be obtained by introducing the correction value d (N−1) for suppressing the fluctuation of the machining amount.
[0077]
[Second Embodiment]
In the first embodiment described above, in the first (first) processing in the continuous processing, it is necessary to keep the sensor 17 in contact with the semiconductor wafer 12 at all times during processing in order to detect processing completion. is there. Even if this is not a problem in practice, the damage to the sensor contacts 17A and 17B cannot be made zero. Therefore, in this embodiment, an improvement is adopted in which the completion of processing at the first time can be detected from the feed position value X (t), thereby making the contact of the sensor 17 during processing completely unnecessary. This improvement makes it possible to use a non-contact sensor (for example, an optical sensor) that causes a problem in the measurement accuracy of the grinding fluid during processing. In addition, this improvement should just be used by control of the 1st process, and the control similar to the first embodiment mentioned above can be performed in the process after the 2nd.
[0078]
FIG. 7 is a diagram illustrating a first control method according to this improvement. This improvement will be described below with reference to FIG.
[0079]
That is, before the start of the first processing, the thickness or surface position of the semiconductor wafer 12 is measured by the sensor 17, and the sensor measurement value S (t) at that time is the thickness Sa (1) before the first processing start. Remember as. Thereafter, the grinding wheel feeding for the first processing is started in a state in which the contact of the sensor 17 is lost, and when the grinding stone 13 contacts the semiconductor wafer 12 by the method described later (that is, the grinding processing start time). Is detected. The feed position value X (t) at this detection time is stored as the first machining start position Xa (1). Based on these values, the first machining completion position Xf (1) is determined by the following equation.
[0080]
Xf (1) = Xa (1) −L−m (Formula 20)
Here, L is the first target processing amount (grinding depth). For example, when the processing purpose is to set the thickness of the semiconductor wafer 12 to a constant value Tk,
L = Sa (1) −Tk (Formula 21)
For example, when the processing purpose is to remove a certain depth Dk from the original surface of the semiconductor wafer 12,
L = Dk (Formula 22)
It can be. Further, as already described, the value m is an estimated value of the grinding wheel wear amount (wear amount at an intermediate point from the start to the completion of machining) in one machining cycle.
[0081]
During the first machining, the machining is completed when the feed position value X (t) reaches the machining completion position Xf (1) obtained by Expression 20. Then, after the first machining is completed, the first machining is performed in the same manner as the method for obtaining the second and subsequent machining completion point value pairs (S (N), X (N)) in the first embodiment described above. Obtain a completion point value pair (S (1), X (1)). Thereafter, the machining control proceeds in the same manner as in the first embodiment described above.
[0082]
As a method for detecting that the grindstone 13 has come into contact with the semiconductor wafer 12 (that is, the start of grinding), there is a method in which a sensor for detecting the contact is separately provided. For example, there is a method of using a sensor for detecting the transparency of the grinding fluid to monitor the transparency before and after the start of processing and detecting contact (start of processing) from the change in transparency. Alternatively, for example, there is a method of detecting contact (start of processing) from a change in electrical conductivity or mechanical vibration due to contact using a sensor that detects electrical conductivity or mechanical vibration.
[0083]
Further, as one method of detecting the contact (start of grinding) without using a separate sensor, there is a method of using a change in current of a motor for rotating the grindstone 13 or the table 11. This is because when the grindstone 13 comes into contact with the semiconductor wafer 12 or is substantially in contact with the grinding liquid, the load on the motor increases. FIG. 8 illustrates this method.
[0084]
As shown in FIG. 8, when the grindstone 13 comes to a position very close to the semiconductor wafer 12 (position slightly before the complete contact with the semiconductor wafer 12), the motor current suddenly rises. This is because the grindstone 13 is in contact with the grinding fluid on the surface of the semiconductor wafer 12 and a load is applied to the motor through this grinding fluid. According to experiments conducted by the inventors, it has been confirmed that the position Xi where the rising amount of the motor current becomes a certain threshold value has high-precision reproducibility. When a certain distance p is advanced from the current rising position Xi, the grindstone 13 comes into full contact with the semiconductor wafer 12. That is
Xa = Xi−p (Formula 23)
It is. Substituting Equation 23 into Equation 20 above,
Xf (1) = Xi (1) −L−m−p (Formula 24)
It becomes.
[0085]
By utilizing this fact, machining control can be performed in the first machining as follows. That is, referring to FIG. 7, the grindstone 13 is basically sent using a standard machining pattern PP_std (thick solid line graph in FIG. 8). However, the idle running position and machining completion position are standard. The following control is performed without following the processing pattern PP_std.
[0086]
First, while comparing the standard machining pattern PP_std and the feed position value X (t), the feed speed Vp is controlled to the rapid feed speed Vpa and then to the buffer feed speed Vpb along the standard machining pattern PP_std. While the grindstone 13 is fed at the buffer feed speed Vpb, the motor current is monitored to determine whether or not the current rising position Xi shown in FIG. 8 has been reached. This current rising position Xi means that the grindstone 13 has reached a close position that only leaves a distance p with respect to the surface of the semiconductor wafer 13. Therefore, as shown in the thick dotted line graph in FIG. 8, ignoring the idle start position Xpa_std of the standard machining pattern PP_std, the feed continues at the buffer feed speed Vpb until the current rise position Xi is reached, and the current rise When the position Xi is reached, the feed speed Vp is reduced to the roughing speed Vpc (that is, a low safe speed to avoid impact contact). Further, the first machining completion position Xf (1) is determined from the current rising position Xi using the above equation 24. At this time, as the value (m + p) in Expression 24, a fixed value set in advance based on the experimental result or the like can be used, or a value obtained by inverse calculation from the past processing result is used. Also good.
[0087]
After the grindstone 13 contacts the semiconductor wafer 12 and machining is started, the standard machining pattern PP_std is compared with the feed position value X (t), thereby switching from the rough machining speed Vpc to the finishing machining speed Vpd. To control. After switching to the finishing processing speed Vpd, the processing completion position Xps_std of the standard processing pattern PP_std is ignored, and the feed position value X (t) coincides with the processing completion position Xf (1) obtained by the above equation 24 The processing is completed.
[0088]
In the second and subsequent processing, the same control method as in the first embodiment can be employed. However, as a modified example, regarding the idling start position, the current rising position Xi may be detected every time ignoring that of the machining pattern PP (N), and the feed speed may be lowered to the safe speed at that timing. .
[0089]
As mentioned above, although embodiment of this invention was described, this is an illustration for description of this invention, and is not the meaning which limits the scope of the present invention only to this embodiment. Therefore, the present invention can be implemented in various other forms without departing from the gist thereof.
[0090]
【The invention's effect】
According to the present invention, sensor contact damage is reduced, the life of the sensor contact is extended, product damage is reduced, and yield is improved.
[0091]
In addition, when an improvement such as that adopted in the second embodiment is introduced, a non-contact sensor (for example, an optical sensor) that is not suitable for use during processing can also be used.
[0092]
In addition, when the machining pattern correction method as described with respect to the first embodiment is used, the entire work time can be shortened and the production efficiency can be improved.
[Brief description of the drawings]
FIG. 1 is a schematic diagram showing a schematic configuration of a main part of a surface grinding apparatus according to a first embodiment of the present invention.
FIG. 2 is a diagram showing the position of a sensor contact at different work stages in the first embodiment, and FIG. 2 (A) shows that before starting the grinding wheel feed and after completion of machining, FIG. 2 (B) Indicates that the wheel is being fed and during grinding.
FIG. 3 is a diagram illustrating a basic control method for feeding a grindstone in the first embodiment.
FIG. 4 is a flowchart showing a processing control procedure in the first embodiment.
FIG. 5 is an explanatory diagram of how to correct a machining pattern PP in the first embodiment.
FIG. 6 is an explanatory diagram of an effect of gain adjustment for suppressing a variation in machining amount.
FIG. 7 is an explanatory diagram for improving the first machining control in the second embodiment.
FIG. 8 is an explanatory diagram of a method for detecting a machining start from a change in motor load current in the second embodiment.
[Explanation of symbols]
10: Surface grinding machine
11: Table
12: Workpiece (semiconductor wafer)
13: Whetstone
14: Whetstone feed mechanism
15: Whetstone feed shaft
16: Feed shaft drive device
17: Thickness sensor
17A, 17B: sensor contact
X: Feed position value
S: Sensor measurement value
PP: Processing pattern
Lfr: free running distance
Xpa: Start position
Xps: Processing end position
Sa: Sensor measurement value of workpiece (semiconductor wafer) before processing
Mofs: Feed axis origin offset
Sofs: Sensor origin offset
X (0), S (0): Origin value pair
X (N), S (N): Machining completion point value pair
PP_std: Standard processing pattern
W: Whetstone wear
E: Error amount due to mechanical deformation
D: Error amount due to thickness variation of workpiece (semiconductor wafer)

Claims (8)

In the apparatus (10) for grinding the surface of the workpiece (12) with the grindstone (13) while feeding the grindstone (13),
A sensor (17) for measuring a surface position or thickness of the workpiece (12) to generate a sensor measurement value (S);
Feed position detection means (16) for detecting a feed position of the grindstone (13) and generating a feed position value (X);
Control means (16) for controlling the feed speed (Vp) of the grindstone (13);
With
The control means (16)
Generated by the previous machining round based on the sensor measurement value (S) and feed position value (X) obtained at the completion of machining of the previous machining round in each machining round when multiple machining rounds are continuous Means (105) for determining a correction amount (Δ (N)) related to the grinding wheel wear amount,
In each machining time, the feed position value (X (t)) acquired in real time during machining and a correction amount (Δ (N) related to the grinding wheel wear amount generated up to the previous machining time. )) And means (111) for determining an estimated value (s (t)) of the surface position or thickness of the workpiece (12) being processed,
Means (111) for controlling the feed rate (Vp) of the grinding wheel (13) being processed and the completion of processing based on the estimated value (s (t)) of the surface position or thickness at each processing time A grinding apparatus characterized by comprising: The sensor (17) is a contact type sensor,
The grinding apparatus according to claim 1, wherein the sensor (17) is not in contact with the workpiece (12) during processing.   The grinding apparatus according to claim 1, wherein the sensor (17) is a non-contact type sensor.   The control device (16) further includes means for acquiring an estimated value (m) of the grindstone wear amount generated during the current machining, and further adds the estimated value (m) of the grindstone wear amount to the grindstone (13 The grinding apparatus according to claim 1, wherein the feed speed (Vp) is controlled.   The control device (16) further includes means for obtaining a correction amount (d (N-1) or k · d (N-1)) for suppressing variation correction of the machining amount, and the correction amount (d The grinding apparatus according to claim 1, wherein (N-1) or k · d (N-1)) is further added to control the feed speed (Vp) of the grindstone (13). Proximity detecting means for detecting that the grindstone (13) is close to the workpiece (12) while the grindstone (13) is being fed,
The grinding apparatus according to claim 1, wherein the control means (16) controls a position (Xf) for completing the machining based on a detection result of the proximity detecting means. Proximity detecting means for detecting that the grindstone (13) is close to the workpiece (12) while the grindstone (13) is being fed,
The grinding according to claim 1, wherein the control means (16) controls the timing at which the feed speed (Vp) is lowered to a safe speed to avoid impact contact based on the detection result of the proximity detection means. Processing equipment. In the method of grinding the surface of the workpiece (12) with the grindstone (13) while feeding the grindstone (13),
Measuring the surface position or thickness of the workpiece (12) to obtain a sensor measurement value (S);
Detecting a feed position of the grindstone (13) to obtain a feed position value (X);
Generated by the previous machining round based on the sensor measurement value (S) and feed position value (X) obtained at the completion of machining of the previous machining round in each machining round when multiple machining rounds are continuous Determining a correction amount (Δ (N)) related to the grinding wheel wear amount,
In each machining time, the feed position value (X (t)) acquired in real time during machining and a correction amount (Δ (N) related to the grinding wheel wear amount generated up to the previous machining time. )) And obtaining an estimated value (s (t)) of the surface position or thickness of the workpiece (12) being processed, (111);
A step (111) of controlling the feed speed (Vp) of the grinding wheel (13) being processed and the time of completion of machining based on the estimated value (s (t)) of the surface position or thickness at each machining time. A grinding method characterized by comprising:

Images