Subscriber Discussion Forum:
FabTime Newsletter, Volume 27, No 2

At one of our customer sites, a maintenance manager requested a new metric: scheduled downtime / (scheduled + unscheduled downtime). The idea is to drive this metric to 100%. This metric would be reported as 100% when the denominator is zero.

This yet-to-be-named metric seems similar to M-Ratio, which (as used in the semiconductor industry) is the ratio of scheduled downtime to unscheduled downtime. A high M-Ratio (≥ 4.0) is considered desirable, meaning that most downtime is preventative rather than unplanned.

M-Ratio approaches infinity as the quantity of unscheduled downtime gets very small. The scheduled maintenance ratio proposed by our customer seems more intuitive to track on an ongoing basis, since it calls for targeting 100% instead of “some large value, ideally greater than four.”

We are considering adding both the new scheduled downtime / total downtime metric and M-Ratio as columns on the Tool State trend and pareto charts in FabTime. This would enable end users to quickly create charts to track these metrics via the Edit Chart capability in FabTime.

Do you think that this new scheduled to total downtime ratio would be useful and if so, what do you think it should be called? We are considering PM-Ratio. Do you favor using this new PM-Ratio or M-Ratio, or both?  Share your thoughts here.

The AG-Grid data table associated with each chart in FabTime offers a robust set of values that can be used to sort or filter the data displayed on the chart. In some cases, especially where customers have extensive custom attributes, these data tables can have dozens of columns. While AG-Grid has considerable native functionality, we wanted to share a quick tip, courtesy of Elaine Jacobson, our recently retired and much missed Program Manager of Customer Success, for finding columns of interest. Simply use the Windows Ctrl+F keyboard shortcut, just as you might on any webpage, to search for your target.

For example, suppose you are looking at a WIP Lot List chart, and you would like to sort it by the Route Family. A quick Ctrl+F for “route f” will cause your browser to highlight the “Route Family” column, as shown below. You can then click in the column header to sort by that column.

You can also search for values within the data table and/or the displayed chart axis labels. In the above example, we could search for a Lot ID or for a Hold Code. We hope that FabTime software users find this tip useful.

A common goal of wafer fab managers is a relatively balanced line. This is not to say that each tool needs to have the same quantity of WIP waiting at all times. There will usually be a few constraint tools that carry higher levels of WIP, on average. Rather, the goal of line balance, in steady state, is to have WIP spread out along each process flow vs. bunched up at steps early or late in the flow, We also want to avoid having all of the WIP at one tool type, of course, but with an acceptance of certain bottlenecks carrying more WIP than other tools.

To generate a Line Balance chart in FabTime, start with either the WIP Pareto chart or the WIP Stacked Pareto chart. In either case, set the “Slice” variable to whatever attribute your fab uses to break up the process flows into linear groups of steps. This may be called Segment, Facility Segment, Sub-Segment, Route Segment, or something else selected by your site. The idea is that the bucket representing each column consists of roughly the same number of steps. Often, week-long buckets are used as Segments, with optional smaller Sub-Segment buckets. It may be appropriate to filter this chart at your site by process flow, or to exclude certain types of WIP (monitors, engineering lots, etc.).

If you are using the WIP Stacked Pareto chart, you can also select a meaningful “Cross” variable for stacking. WIP state, product family, and module are commonly used.

Optionally, use the “Stripe” option in the Formatting section below the chart to add a line that represents the average WIP across all included segments. This is what the WIP would be by segment if the line was perfectly balanced. An example is shown below, stacked by WIP state. We see that segments 4 and 6 have excess WIP, while segment 5 is starved. The stripe value was calculated by dividing the total WIP (shown in the chart header) by nine (the number of segments).

What buckets do you use for the x-axis when you look at line balance? What do you stack or filter this chart by at your site? Share your thoughts here.

A subscriber and cycle time class participant wrote to say that there was debate within her organization about the correct definition for Open Lot Cycle Time. Possible definitions being considered by the site included:

• The time it takes to get from one identified step to a later identified step. For example, the time it takes to get from step 1500.1200 to 3001.1000, which could represent a specific module.
• The number of steps completed in a certain time frame. For example, how many steps were completed from Jan 1, 2026 to Jan 15, 2026.
• In the case of open lot DPML calculation, average the number of mask levels completed for all lots over a period of time.
• And the latest from an AI chat engine is the time it takes for a specific lot to go from start of line to fab out minus all the time spent on hold.

Response from Jennifer: In our FabTime reporting module, Open Lot Cycle Time is defined as follows:

1. Pull all records that match filter and Lot has already started.
2. Compute current factory age = AsOfTime - factory start time.
3. Compute remaining planned cycle time based on RTG_ROUTE_STEPS_PLUS.TARGET_STEP_SEC_TO_EOL 
   1. If Lot is currently on hold or rework, find last NonRework record. 
   2. If step not found or if step is prior to start time, remove the lot from list
4. Compute Open Lot Cycle Time = (Age from start to last Move) + remaining planned cycle time.

This is essentially current CT + remaining planned CT, calculated for each lot individually and then aggregated as needed. Other metrics for estimating future cycle time based on current performance in FabTime include dynamic cycle time (DCT) and dynamic x-factor (DXF). See Forward-Looking Cycle Time Metrics (Issue 24.03) for definitions of DCT and DXF.

In our FPS Cycle Time Analyzer, open lot (AKA dynamic) cycle time is calculated differently. For each route step, we compute a weighted average cycle time based on moves for the route step combination over the selected time period. For a given route, this is summed across all the steps for that route. A weighted average across routes is computed for the factory. Fill-in logic is used if there have not been enough moves completed at the route-step level.

There are benefits to each of these methods for different purposes. Readers, how does your fab define Open Lot Cycle Time?  Share your thoughts here.

While discussing open lot cycle time metrics, the customer cited above added this comment on dynamic x-factor:

“I haven’t found dynamic x-factor to be an insightful metric for us because it doesn’t correlate well with improved productivity. I believe this is because we have an inconsistent amount of available WIP. When holds are high, DXF is high because there’s not much WIP to keep track of. But tool loading is low because there’s not much WIP. We’ve found the raw number of wafers loaded to be a better metric than DXF”.

Response from Jennifer: This point arose recently at a customer site that currently has low levels of WIP running.  If the quantity of WIP running on tools frequently drops to a very low value, this will drive spikes in the DXF calculation (DXF = Total WIP / WIP currently running on tools). It seems to me that using raw number of wafers loaded instead of using DXF makes sense in this situation. Raw number of wafers loaded won’t give you a forward look at cycle time, but it will tell you if your fab is doing a good job keeping tools busy. This metric will, of course, drop down at shift change if your fab isn’t doing a good job keeping tools running at that time. Do any other subscribers have real-world experiences to share with using DXF or other forward-looking cycle time metrics?  Share your thoughts here.

Past issues of the newsletter are available for subscribers to download in PDF format. Existing subscribers can find the archive link in your most recent email newsletter. New subscribers will see the link upon registering. 

• Register for the newsletter here. 
• Main article for Volume 27, No. 2: Setup Time, Tool Qualification, and Cycle Time
• FabTime Newsletter Community Announcements Page