Thread: Identifying & Controlling Errors

From the K3 flip stop thread, Its a long one but well worth the read! (I hope!)

Derek must have had a long day when he mentioned 3 degrees and not 0.3 degrees as shown on his Wixey gauge or similar.

If the flip stop was out 3 degrees one would have discovered it long ago, we would hope! 3 degrees would equate to 5 mm out of perpendicular in 100 mm.

0.3 degrees is 0.5 mm in 100 mm so still significant for fine joinery, or approx 0.1 mm on opposing faces or shoulders of a 19mm wide board / tenon.

My surveying background comes of benefit at times especially one maxim "work from the whole to the part" and our obsession with controlling errors within accepted tolerances.

How does this apply to a flip stop on a saw??? Read on as this may assist in understanding how to limit errors in wood working.

"Working from the whole to the part" is critical in containing or limiting errors.
Knowing what the source/s of those errors are, is also critical.

Example - for a complex wood working component it is simply adopting a process of marking the whole length in one measurement. QA / QC'ing that each main component length is correct, then marking key features, say a mortice, which then becomes a key "reference point".

Lesser / secondary features such as say dowel holes etc are then set out from those "reference points." Such a process limits the effect of "errors."

Use a system of set out that references all measurements from one or a small number of datum points (the lesser the better) which have been meticulously set out and checked. Using multiple "datum points" can be beneficial when there are logical groupings of components, say shelf pin holes.

Don't use a leapfrog measuring technique, as it compounds errors very quickly, where an overall or "progressive" measurement is made using several individually marked measurements to arrive at that (total) measurement.

If one were to set out say 6 items along a board in individual measurements, there are potentially 12 sources of error - a zero or datum offset error & a marking or "set out" error for each measurement.

So for ease of illustration if the datum and set out errors were say 0.5 mm each then that equates to an "error circle/ellipse" of 0.5 mm per individual measurement bay, but some -+ 1.73 mm (standard deviation) for the 6th point set out.

However if every component / feature was set out by one "whole" measurement from a datum reference it would still be only - + 0.5 mm for each item, so the sixth point or the first to last point would all have a potential error of +-0.5 mm, and not +-1.73 mm for the 6th. Makes for a very significant improvement in precision and accuracy.

Now remember that bit about knowing what are the potential "sources of error?"

In the above we were only discussing the "measurement errors" associated with setting out, but have not considered any other sources of error - yet!

Errors, yes we all make them, the good operators find them first, before someone else does!

Errors tend to be random, so they are likely to have as much probability of being "under or over" UNLESS we have introduced a "systematic error" - an error that will always be there - say a repaired tape end that is no longer "zero." If we know that the error exists we can accommodate for it. If we are totally oblivious to it then we blissfully keep making the same mistakes!

All good surveyors, (and wood workers) "calibrate" their "measuring tools" so that they understand the potential sources of error that may affect their measurements.

This effectively involves checking that the measuring device has been manufactured to and complies with an "accepted measurement standard" to within accepted tolerances (actually required by legislation for surveyors.)

Not a problem if you only use one tape measure and everything stays in-house! However if you supply components / items to others its a good idea to make sure that your "your one meter" = "the customers one meter."

We generally accept that our, and others, tape measures, rulers or squares etc comply to our "Australian Standards" but that may not be the case due to poor manufacturing quality control, or fair wear & tear.

So, for say a typical retracting tape measure there is the "zero or offset error" - is "zero" actually "zero?" There is also a "scale error" - is one meter actually "one meter" as described in legislation / standards?? Is the accuracy of markings consistent for the whole length of the tape, no missing graduations??

We can't check everything or to that level of detail, however we can check some simple things.

Very carefully set out say 300 mm "reference measurement" from one end of a squared board using a "quality steel rule" (your "reference standard''). Then measure that "reference measurement" with your tape measure using the "tab end" end in both compression & extension modes, and using a number of random "bays" along the tape measures length, say 500 to 800 mm, 1.2 to 1.5 m then 2.8 to 3.1m, 5.3 to 5.6m etc.

All measurements should fall within a statistically valid range, for fine wood work certainly around 0.2 mm or less.

Now a similar process applies for all "measuring tools" used to "measure" or "check" the measurements we make, steel rules, tape measures, vernier calipers etc; and includes measures on saw table fences, flip stops, etc.

Each "device" is a potential source of error, plus they may not all "measure the same".

So after a long winded explanation - back to the flip stop & measuring fence!

The "measure" on the fence has a "zero error" which may in fact be dependent on the saw blade (kerf) being used. A 1.8 mm wide kerf vs a 3 mm wide kerf introduces a potential "zero or datum error" of anything from zero to 1.2 mm depending upon which side / face or even centreline of the saw blade the "measure" was originally "referenced" to, and to which side of the saw blade the measurements are taken. It may well be a different "zero / offset error" for L and R for each saw blade!

Next is the "flip stop" - or in the K3 specific example a flip stop that has a 0.3 degree error in the vertical, perpendicular to the saw table! Now, it potentially may also have an error in the horizontal, perpendicular to the fence!!!

We will only consider the one error for now.

If the fence is 100 mm high AND the measure is on top of the fence and it has been accurately referenced as "zero" to the face of the saw blade then a 100 mm thick / high stock item could be cut to say an accuracy / precision (they are different) of say 0.2mm, or better if you are OCD.

However if the stock is only a 20 mm thick board that "error" actually becomes a combination of two "errors" - one the "zero / datum error" associated with the "measure" - + 0.2 mm; and the "flip stop perpendicular error" which is - + 0.4 mm (0.5 / 100) *20 = 0.4 mm. So the "error circle" is around - + 0.44 mm for 19 mm stock; 0.32 mm for 50 mm stock, and - + 0.2 mm for the 100mm stock etc.

As Chris has experienced with the errant flip stop; that "error" can actually be different for each side of a tenon because the "stock" effectively changes "thickness" when the waste is removed and the component "flipped" then "butted up" to the stop, to complete the second cut.

Now back to working from the whole to the part!!

Quality assurance / control, is key to successfully "measuring" and "cutting" components with precision and accuracy.

Using test pieces (of similar dimensions) at each key step in the process to QA say setting up to cut a series of tenons is sound practice. Perform the setting up, cut, then check measure the test item.

When the item meets your expectations of “quality” go ahead and cut the real stuff!

Then there are the “labelling errors” and “cutting errors.”

Now, if you cut “wrong side of the line” because you didn’t effectively “mark out” – well, we’ve all done that!
Plus, minus one kerf width!

Minus! Damn! Double damn!!