Applied Mechanics and Materials Vol. 421 (2013) pp 898-903 © (2013) Trans Tech Publications, Switzerland doi:10.4028/www.scientific.net/AMM.421.898
Algorithm to Improve ProcessRobustnessin the Assembly & Test Manufacturing Industry A Case Study of the 1064nm Wavelength Laser Mark Equipment Darin Moreira Anthony Vincent1,2,a, Bhuvenesh Rajamony2,b 1
Finish Engineering, Intel Technology (M) Sdn.Bhd, Penang, Malaysia;
2
School of Manufacturing Engineering, University Malaysia Perlis UNIMAP, Perlis, MALAYSIA a
[email protected] ;
[email protected]
Keywords:Maintenance algorithm, Laser mark, IR laser, TRIZ, Quality output, Case study
Abstract.This document explains and demonstrates the generic improvement algorithm created to enhance the maintenance methodology in the semiconductor manufacturing environment. The use of this algorithm demonstrates how a process and equipment can utilize it and get better output quality from a process and cost standpoint, which is a key driver in any manufacturing industry. Introduction In any Assembly & Test Manufacturing (ATM) factory/company, output is the key and the more output that is generated the more money the company makes. Having output alone is not sufficient as it needs to be of the best quality to ensure zero customer returns or complaints as this result in bad future sales. To achieve Quality Output, some aspects that need to be taken into account and is not limited to the following, cycle time, equipment performance, operator attentiveness to details, process yield (rejects per process), process controls , etc. All these factors play a major role in achieving quality output but for this particular paper, we will focus on the equipment performance as the factor to influence the quality as well as identify ways to improve the manufacturing process through specific monitoring that eventually improves the maintenance policy of the equipment. We will be using the laser marking tool as the case study for the paper. Laser Mark Process background Marking is an integral part of the manufacturing process for any semiconductor company that manufactures computer processors, chipset and even server based products. Marking done on the completed units carries a great deal of information that is very important to the end customer that is purchasing these parts. The markingtells the customer the product specification for example the product type, its speed, country of origin, year of manufacture, batch number and much more. Therefore a unit without marking or illegible mark becomes meaningless to the customer. In conclusion, the marking quality is critical to the overall manufacturing process. Laser marking is widely used in Intel Corporation to mark all the products that it manufactures worldwide.To ensure the laser tool performs to its best capability, the tool needs to be maintained well to produce output. Due to the complexity of the tool and high demands of production, toolstability is critical. 1064nm wavelength laser (IR laser)[1]. The 1064nm wavelength laser (infrared rays on the light spectrum) is an industrial laser that is used for marking in many manufacturing plants around the world. Its function is to engrave the marking information required for a specific product. The different marking surfaces require different tool output (Power) for example marking on the integrated heat spreader (IHS) will require higher power compared to marking on the Ink Swatch that uses a fraction of the power to mark compared to the IHS. The difference in power used determines the stress the diode has to deal with thatinfluences the capability of the diode over time. High power usage shortens the diode lifespan and likewise lower power usage slows down the degradations of the diode.One of the main design features of the 1064nm IR laser is the resonator that is able to generate high power to mark various products. However to ensure we are able to get max power out of the diode, perfect All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of TTP, www.ttp.net. (ID: 1.9.65.122-05/09/13,10:52:48)
Applied Mechanics and Materials Vol. 421
899
alignment is needed at all times during the laser lasing [2]process (refer Fig.1). This coupled with the beam alignment when the beam is split in two to provide equal amount of power to both Galvo heads that controls the marking on the surface of mark.The biggest challenge is to ensure alignment stays perfect as the most expensive part in the laser tool; the Diode, Nd:YAG[3]rod and Quality Switch rely onalignment to work properly and will degrade when the alignment is off.
Figure1: Diagram of a basic laser resonator Maintenance Practice in Intel At Intel (in general), Preventive Maintenance (PM) is the maintenance methodology that is used widely across most of the equipment in the ATM network of factories globally. Typically the PM is broken down to different types by its timeline i.e. weekly, monthly, quarterly, semi-annual and annual (refer to Fig.2). The differences in duration also indicate that there are differences in the parts changed and timeline required to perform these changes to the tool during each PM type. PMs are classified under scheduled downtime and the equipment is down periodically according to the PM type to carry out the changes and to service the tool.
Figure 2: Preventive Maintenance standard durations for Laser Mark tool Downtime is classified as unproductive time by the equipment due to a failing part/broken connection that prevents the equipment from functioning as normal. Business process in the semiconductor industry operates at 24x7 so when a downtime occurs, the cycle time of the process gets impacted withzero output. This is an undesirable state and if the root cause of the downtime is related to parts breaking down, then the PM is deemed ineffective.For a process to be efficient, scheduled downtime (which is planned and predictable) should minimize the occurrence of unscheduled downtime. PM and engineering activities are all classified as scheduled downtime and if performed well (PM personnel competent, parts used in good condition and PM process effective) then the unscheduled downtime would be minimal.From the downtime database (refer to Fig.3), we found that around 2/3 of the downtime was related to defective parts, parts wear andtear and insufficient process control (Blur Mark 44% + Out of Position20% + Illegible Mark 2%).
Figure 3: Excursion Downtime 2009-2012
900
Information Technology for Manufacturing Systems IV
Concerns & Problem Statement: The mark process in the ATMpipeline is done at a stage where the products are fully packaged and have 100% passed Test and is only left with the laser mark, visual inspection and packing prior to it being shipped out. Any failures or rejects after Test is very costly at this point as the complete cost of building the unit is wasted. At the laser mark tool, the cycle time for each batch of material is very short due to the nature of the marking process that is done very fast. For this fact, there are a few tools only required within a factory to be used to mark hundreds of thousands of units a week. If there is a tool down, then the whole pipeline will be impacted due to the pile up of material at laser mark process, therefore it is critical that the tool is always up. Problems are broken up into 2 i.e. bad quality escapees and process failures. We will not discuss bad quality failure in this paper but focus onprocessfailures.Forprocess failures, laser mark contributes by having no mark on the units, parts broken causing intermittent marking quality deviation or even complete breakdown in tool like diode failure, Q-switch broken, etc.. All these types of failure cause units to be rejected. Algorithm to enhance PM effectiveness in the manufacturing industry When we start to ask the question why all these failures happen, we eventually reach the fundamental root cause that there are no detection of any sort for this to be known prior to it occurring. The fact also states that the PM done isinsufficient to prevent these kinds of problems from happening altogether. Therefore there is a need to improve the maintenance situation as well as the process. Per the General Maintenance decision making process based upon output based maintenance (OBM)(refer to Fig.4) [4], we further derived the “Take some appropriate action” step by putting in specific “mini steps” to create a new algorithm to tackle manufacturing related problems thus the general improvement flow was further derived to become the Algorithm for Process Improvement in the semiconductor industry.
Figure 4: General Improvement Flow This algorithm (refer to Fig.5) proposed uses asystematicapproachto improve the overall equipment performance and process robustness as compared to solely depending on the PM practice to sustain the equipment. In our case study, we are proposing the use of the algorithm to enhance the effectiveness of the overall process of the laser mark tool as the current PM is insufficient to minimize unscheduled downtime at the same time prevents quality issues from occurring.
Applied Mechanics and Materials Vol. 421
901
Figure 5: Algorithm for Process Improvement in the semiconductor industry Part 1: Defining the MPV. The Main Parameter of Value (MPV) is the critical features and functionality that differentiate products and drive customer purchasing decisions[5]. In manufacturing terms, this can be reiterated as the critical feature and functionality that the process is dependent on and directly influences the tools performance. The best waysto derive the MPV in manufacturing arefrom:1) Process, 2) Equipment and 3)Cost impact.These 3 aspects outline what the MPV should be and use the Cause and Effects chain analysisas well as the Function model to clearly identify the MPV.For process related segment, the quality of mark can be monitored as there is a post mark camerathat is used on the tool. As for the equipment related segment, the parts like Diode, Nd:YAGRodand Quality Switch are the important parts that go hand in hand with the cost as these are relatively the most expensive consumable parts in the laser tool and at the same time there is no way to tell the condition of these parts when it is being used (current situation).In this segment, we employedTRIZ[6] to bring the MPV to the next level (this section is not limited to any specific problem solving tooland it is user dependent). In this project, we used TRIZ Contradiction Matrix[7]with the 40 Inventive Principles [6]to achieve our goalsas we came across a contradiction when we wanted to improve the engineering system.ImprovingParameter:Accuracy of Measurementand Worsening Parameter:Complexity of device.Based upon the recommendations that we validated, we found that Principle #10: Preliminary Action is the best to improve the detection capabilityby utilizing the resources that is readily available.We identified the parts that impact our MPV (Diode, Nd:YAGRod and Quality Switch) and identifiedhow we want to measure this change/degradation. Part 2: Quantifying the Improvement (Measuring& Tracking the MPV).For the Diode and Q-Switch, these are parts that are not measured at all during the process but these are critical parts in the laser operation. The diode’s functions as the heart of the laser tool, as it transfer current to excite Nd:YAGcrystal to emit light that resonates and produces laser. Without the diode there is no laser. There is no measurement method for the Diode in the industry at the moment. The Q switch on the other hand functions to stop the lasing process by refracting the laser beam. It acts as an electrical switch to with the use of radio frequency to effectively bend the laser. The effectiveness of the Q Switch is only measured during the PM but if there is a fault with the part, it will show immediate signs on the marking. We need to device a new method to get this done to achieve our goals per Principle #10 (Preliminary Action).For the Diode, part of the replacement process when a new diode is fitted into the tool is that the diode is aligned to firstly provide maximum power. Assumptions made here is that the mirrors rear (fully reflective) and front (partial reflective) and all other parts inside the resonator are in good condition. The power is then measured at specific current ranges from min to max to ensure power output is at the peak. The final check point is a measurement method called the 20mm/30mm where the beam position is measured using a specific fixture that measures the horizontal and vertical position of the beam after alignment is done. It must meet the 20mm/30mm
902
Information Technology for Manufacturing Systems IV
targets or else the alignment is out and will impact the splitting of the beam that directly impact the marking quality. This process of changing the Diode and performing laser alignment can only be done by trainedpersonnel or by the tool manufacturer’s Field Service Engineers and the process to get this done takes approx. 6 hours. Assumptions to measure condition of the diode without opening the resonator rail: 1. Assume the laser alignment is still in good condition (Spec:Able to achieve20mm/30mm) a. Measure the output from Galvo Head 1 and 2 (Spec: Must be 70 andBackward
