Info
Info
News Article

Cost-of-ownership As A Real-time Control Metric

The earlier a cost issue is identified, the fewer wafers will be impacted, the less processing time wasted, and the lower the cost, write Howard Ignatius of Electroglas' EGsoft division and Daren Dance of Wright Williams & Kelly. Traditional static COO analyses help in making purchase decisions, but require more frequent data collection for process control. When real-time data are used to calculate COO and the result is analysed with proven SPC techniques, you have a tool to dynamically monitor and control the cost of a process




The key to success for a chip manufacturer running 300mm silicon is overall fab efficiency (OFE), an extension of the traditional concept of overall equipment efficiency (OEE). OEE, in turn, is a sub-set of cost-of-ownership (COO). Parameters for COO, such as tool throughput and utilisation, determine manufacturing efficiency. But traditional manual data collection has limitations: the frequency of new data points can be insufficient to drive real-time decision processes. Traditional data collection techniques are inappropriate for continuous process improvement. If COO data are gathered in real time, it becomes a dynamic basis for decision-making. Not only can it identify why a "bad" tool is bad, but it can determine why the "good" tool is better. The use of dynamic, real-time COO can be the fast track to optimised OEE.


The inputs to COO calculations (Figure 1) such as tool throughput and utilisation are the same parameters that on a fab scale determine manufacturing efficiency. Combined with statistical process control (SPC) methods, dynamic COO can be used to:


* Monitor and continuously update assumptions

* Report progress toward cost control

* Optimise the process based on best-known, rather than theoretical, return on investment (ROI)

* Make informed cost-reduction changes

* Convert simulations and forecasts into measurable process control tools


Real-time COO

Using real-time COO, a change in an input parameter can be evaluated in terms of its effect on the components of COO. Figure 2 illustrates the concept using training as a variable. If more money is invested in the training of maintenance and operating personnel, the real-time COO will reflect this in other cost metrics such as equipment availability (improved uptime), labour (more efficient operation), and maintenance (lower mean time to repair).


To illustrate the use and benefits of real-time COO, we examine examples from two very different processes, wafer sort and tungsten chemical vapour deposition (CVD). The concept can be extended to any tool that conforms to the SEMI E-10 guideline (Definition and Measurement of Equipment Reliability, Availability, and Maintainability), as well to any combination of fab tools such as a work cell.


Wafer sort

As devices have become more complex, the cost of wafer probe test has increased significantly. Test pads have shrunk to accommodate higher pin counts so probe alignment is more demanding. Voltage and current have decreased, making the control of test parameters more critical. Temperature has become a significant test parameter. Additionally, over the recent past, probe card costs have increased ten times, probe stations have doubled in cost, and test systems have gone up 5-10 times. For these reasons, real-time COO during prober operation is a welcome tool for cost control.


The configuration for real-time COO data collection and communication is illustrated in Figure 3. Each test probe station (left side) is connected by a two-way communications link to a central server. The server, in turn, connects to the corporate intranet and can also be linked outside the site via a secure internet connection. Data and reports can be accessed internally and external via standard web browser software. Customised reports are generated from the prober data and can be sent via intranet connections to users within the company. Proprietary yield data are blocked from the internet by server software-only tool-related information can be sent outside the enterprise. An open data acquisition format allows data to be accessed from any open database compliant (ODBC) test station. The data from the test floor can include reports on the equipment state, wafer yield maps and bin monitoring.


The interrelation of process variables and fixed costs is shown in Figure 4 as a logical summation process. Variables feed into a logical NOR operation (purple), and then together with fixed cost data (capital, facilities, consumables and such), feed into a logical AND operation (green). The web-based output appears on the computer display such as shown in Figure 5.


Figure 5 is an example of real wafer sort data for three probe stations in the form of 10-day SPC charts. Other time intervals could be selected, or, to gain detailed insight, wafer-by-wafer data (cost/wafer) could be tracked. Instead of traditional yield data, such as average number of good die per wafer, the parameter plotted is 'COO', which is derived from a modified version of the COO formula (Figure 1) that includes real-time tool parameters. In this example, each point represents the COO for one day. Note that control limits differ from station to station: control limits are calculated by collecting data from a statistically significant number of wafers for the specific probe station. This makes it easy to determine which probers are operating at less than optimal efficiency. Violations of user-selected SPC rules are flagged as red data points.


Figure 6 shows real wafer sort data for 22 days from each of two probe stations. The bottom chart shows rule violations for days 13 and 14 (more than seven points below the centre line) and for the last day (one point above the upper control limit).


Figure 7 shows how details of an SPC rule violation can be displayed for three probe stations. Wafer sort data are displayed for a sequence of ten days of data from a single probe station. Each point represents the average of all the data from one day. The data for the first and last days have been removed for a special (known) cause and the COO value for those days is displayed. The inset (Data Tip) shows the details from the date (Jan 11) selected by the cursor. In this case it displays the specific SPC rule that was violated.


Tungsten CVD

The advantages of combining COO and SPC allows one to continuously monitor and update COO assumptions. Simulations and forecasts can be converted to measurable process control tools. Process changes can be implemented and the results immediately available that measure the progress towards cost containment goals. The process is optimised based on the best possible rather than the projected return on investment. Process decisions that reduce costs can be validated with timely, continuous data analysis.


Table 1 and Figure 8 show the results from the optimisation of a tungsten CVD process. The expected cost per wafer was about $13. The actual average COO was an unacceptable $21 per wafer and varied widely. The best COO was about $6 showing that the process could, when in control, exceed expectations. This gave an incentive to attempt process optimisation. The optimisation steps included adding a pre-pumpdown purge, increasing pumpdown time, and monitoring the chamber pressure after the plasma clean step. Although the first two changes may have been expected to decrease wafer throughput, the positive changes in other parameters included in the COO outweighed this effect. Without a COO analysis it might have been more difficult to justify a change that might impact throughput. The final average COO equalled the best COO obtained before the process changes and the variability was significantly lower (Figure 8). Continuous COO monitoring should aid in further reducing the variability.


Acknowledgement

This article is based on a presentation given at the Advanced Equipment Control/Advanced Process Control (AEC/APC) Conference, Dresden, Germany, April 10-12, 2002.








NextEnergy Capital Acquires Its First Asset In Portugal 17.4MWp Solar PV Project
Power Roll Trials Solar PV To Power Up Himalayan Villages
Sunstore Solar Launches WattGrid, A New Range Of Turnkey Off-grid Power Systems
Tandem PV Devices Feel The Heat
Low Carbon Develop UK’s Largest Community-owned Solar Park
Solar Power As Rental Offer Launched By Aggreko
Going Green In Lancashire – Hundreds Of Houses Installed With Solar Panels In Ground-breaking Project
Greencoat Renewables Announces First Transaction In Nordic Market
Ingenious Invests In Electric Vehicle Charging Firm
FIMER Powers UK Largest Rooftop Solar Project
UK'S Largest Battery Ready To Balance The Grid
The Smarter E South America Postponed To October 18-20, 2021
Oakapple Renewable Energy Appoint Stuart Gentry To Head Business Development
TLT Advises Innova Energy On £30m Refinancing Of 57 MW Solar Portfolio
TLT Advises Santander On 30MW Flagship Battery Storage Project
New Innovation Set To Change Renewable Energy Market
Habitat Enerdy Enters Balancing Mechanism With Largest Battery
FRV And Harmony Energy To Develop Second UK Utility Scale Battery Project
UK Green Tech Company Myenergi To Double Workforce By 2021
Sharp Launches New 440W Half-cut Cell PV Panel
SOLARWATT Links With Easy Roof To Provide Building-integrated PV For Better-looking Buildings And Smart EV Charging
Everoze Creates Skyray To Design And Engineer Great Solar PV Projects
Analysis Of UK Commercial Roof Space Shows Solar PV Film Can Achieve Net Zero Without Greenfield Sites
Sonnedix Named ESG Global Solar Power Generation Sector Leader By GRESB
×
Search the news archive

To close this popup you can press escape or click the close icon.
Logo
×
Register - Step 1

You may choose to subscribe to the Smart Solar Magazine, the Smart Solar Newsletter, or both. You may also request additional information if required, before submitting your application.


Please subscribe me to:

 

You chose the industry type of "Other"

Please enter the industry that you work in:
Please enter the industry that you work in:
 
X
Info
X
Info
{taasPodcastNotification} Array
Live Event