About measuring sharpness *LINKS* *BS*

A-MEN!

Denis Ch�nard, Orl�ans, Ont.

>

I think there is defensable data out there

bill tindall

>What do you find lacking with the approach Steve Elliot uses to answer these questions? Or Brent Beach?

Re: Great Book Recomendation!

Jim Toews

>

I don't....but I'm often wrong

woodburnbob

>Good reminder, Bill. I wasn't thinking of Steve Elliot or Brent Beach when my comments squirted out. I have the highest regard for Steve, his site and his insight. I've gone there many times over the years to see what he's up to, but it's been a while and I hadn't seen (or remembered) his functional tests with thread cutting.

As with everyone, I had seen Brent's microphotos on wear and then those from Steve some years ago. Those photos do give face valid, qualitative data about wear, given lots of assumptions about the starting point of sharpness, equivalent wood and technique and so forth. But I really don't think those photos or measurements of reflections tell us anything much about original "sharpness". Perhaps, I'm wrong. They might tell us about edge failure or wear in various steels given certain preconditions, but I wouldn't think they'd ever be able to answer questions comparing different sharpening products or techniques.

The thread cutting, which is an attempt at a functional measure of intitial sharpness is a reasonable approach to answer sharpness questions if the measure has the properties of reliability and validity. I did just now study Steve's newest comments, graphs(data) and conclusions. What I took away was 1) he thinks differences in original sharpness are miniscule and don't seem to have much practical meaning (he can get everything very sharp), 2) using the mean of 7 measures per trial gives lots of scatter when plotting these means against best fit regression lines and he's thinking 14 measures per data point may lead to a better mean and standard deviation which might improve the regression fit, 3) thread cutting may lack reliability and validity across varying combinations of steel and abrasive. I hope someone pings Steve to give him a chance to correct all my misperceptions and set the record straight.

So, from what I read, Steve is saying his current approach may be too inaccurate to yield reliable and valid conclusions.

But, Bill, you are correct, Brent, Steve and others out there are working on this. In fact, I was interested to follow one of Brent's links to this site, whose author hopes some day to use CATRA machines toward this very end. So I'd like to see a deep-pockets group fund and equip guys like Steve with big bucks and the right instruments to really get to the bottom of some of the mysteries.

a lengthy discussion of result significance

bill tindall

>"But I really don't think those photos or measurements of reflections tell us anything much about original "sharpness". I agree, the objective here was how fast does the metal wear. Lacking is any replication from which statistical significance can be gleaned, I can't tell if the differences posted are meaningful or not. However, they agree qualitatively with Steve's data.

"Steve : 1) he thinks differences in original sharpness are miniscule and don't seem to have much practical meaning (he can get everything very sharp), 2) using the mean of 7 measures per trial gives lots of scatter when plotting these means against best fit regression lines and he's thinking 14 measures per data point may lead to a better mean and standard deviation which might improve the regression fit, 3) thread cutting may lack reliability and validity across varying combinations of steel and abrasive. " I have had the opportunity to do statistical analysis on Steve's earlier data. He took tons of data to construct a curve, vastly more than needed, in my opinion, to draw statistically significant conclusions comparing blades. He was worried that persons not familiar with data analysis would see a point out of line and dismiss the conclusions. For skeptics (and I don't mean it negatively) plots with 2 sigma error bars on the points would be a welcome addition, but it would only bewilder others. Better yet would be to post the raw data and let persons analyze it if they chose to.

"So, from what I read, Steve is saying his current approach may be too inaccurate to yield reliable and valid conclusions." Steve is cautious. Again, I have not seen recent raw data, but there is no question his ealier data was sufficiently precise to yield significant conclusions about the relative wear and initial sharpness of various steels. The only area that one could argue is the test itself. It seems reasonable to me, and mirrors the commercial test you pointed to in your first note.

The two big questions to address are how does the edge life of different steels compare and how sharp is sharp enough. Steve has answered the first question to my satisfaction, at least for the steel samples he has in hand. I expect heat treating is a more significant source of error than Steve's data. As far as initial sharpness, beyond a point does it matter? I have pestered Steve to answer this question for years now as I think it is the next most important contribution he could make. It may well be that beyond a certain point an edge is so delicate that 10 feet of planing renders it equivalent to a lesser intitially sharp edge. This is the hypothesis I currently believe in. I know that by the time I sharpen a plane iron it is way duller than when I started. Hence, during its useful life it perfoms satisfactorily at a sharpenss vastly less than it can be sharpened to. It remains for Steve to answer the qeustion of whether sharpening to a higher degree will result in longer practical life.

I recently got edge images of a 3V iron fairly crudely sharpened (the objective here was to see what imaging could be done on the first iron I could spare). There is no doubt that CPM 3V can be sharpened to an edge thinner than 1 micron. Diamond mows right through the carbides so grain size places no limitation on the refinement of the edge. I can't imagine that a 0.1 micron edge lasts long in maple, so does it matter whether a steel can be taken there?

So I'd like to see a deep-pockets group fund and equip guys like Steve with big bucks and the right instruments to really get to the bottom of some of the mysteries. " LV has data but they are not revealing it.

And a little more....

woodburnbob

>Thanks for the extra information, Bill. I didn't know you had worked on Steve's data. If it went up on the board, I must have missed it.

He was worried that persons not familiar with data analysis would see a point out of line and dismiss the conclusions.

Kind of depends on how many points are "out of line" and what the conclusions are. I'd rather see all the data and reach my own conclusion.

My real questions about the thread cutting approach are about the stability of the measure: whether humidity and temperature account for anything, whether thread in a spool is consistent, the effect of pretensioning thread, and all those sorts of basics that go into proving a measurement technique actually is reliable and valid before using it to test hypotheses. These are exactly the same things I wonder about regarding the "industry" machines based on cutting card stock via CATRA. For all I know, CATRA reincarnates the Wizard of OZ.

The two big questions to address are how does the edge life of different steels compare and how sharp is sharp enough.

From the point of view of what is the best steel or blade vendor, I suppose you are correct. My list of big questions is much less practical and a bit larger.

I'm curious about which nuggets of received woodworking wisdom, repeated through generations of old guys lecturing young guys, are bogus. For example, do >4000 grit water stones really get a blade sharper than, say, a two-bit 1940s carborundum stone? How much sharper can you get a chisel with a $750 natural stone from the hills of Japan compared with a $20 Norton India stone, or what have you? For sharpness, does it matter whether you push or pull the blade? What, if any, relationship exists between sharpening pressure against a stone and how sharp a blade gets? What really does lubrication contribute to sharpness using oil stones? Can a person with a mammalian nervous system and average coordination avoid dubbing an edge and keep it optimally sharp by free-handing as well as a reptile who uses a sharpening jig? Irreverent and weird questions, I guess. But from what I read, many folks would have a firm but contradictory opinions on these topics. I'd just like to see proof of who's right and who's wrong.

Re: And a little more....

thomd

>"It may well be that beyond a certain point an edge is so delicate that 10 feet of planing renders it equivalent to a lesser intitially sharp edge. This is the hypothesis I currently believe in."

I haven't seen that it seems the sharper it is the longer the edge lasts. I'm not really sure if there is a difference in pure cutting like cutting a carrot and cutitng like planing that is across the edge. So I guess it is an interesting question to resolve.

I don't think pulling a plane makes it sharper, but the plane can remain useful for a much longer time, months past any vestige of real sharpness having left the edge. I supose there is the speed argument; faster equals sharper.

Re: And a little more....

Miles

>I will try to find time to post some of the sharpness / wear data I have that may be interesting. Measuring power used vs cutting angles & speeds has been done extensively for machinery, at least, & I have some of this in wood technology ref books. Maybe some can be applied to hand tool cases too.

The one real question I have over Steve Elliot's thread test (that he rightly is cautious about) concerns the difference between the cutting situations of a 'worn' blade in a plane vs in the test. In a plane it is presenting the worn flat to the wood surface (as it gets larger, the pressure drops and the blade doesn't penetrate). The relief angle adjacent to the edge has gone to zero. This probably means that the edge angle is now 45 degrees (standard frog) where it was 34 degrees when sharpened. The thread test is comparing the cutting force between a 34 deg and a 45 deg edge, both of which are probably quite sharp at the actual edge (normally the wood wears away the steel, rather than fracturing it or making it curl up).

Steve suggests that the thread test may not really reflect the planing behaviour of a worn edge, and I think he is right and this is why. We know that more obtuse edge angles cut perfectly well (in planes, scrapers, and in turning) as long as they are presented to the work correctly - probably meaning that there is enough pressure to penetrate the surface. The wear reduces the effective pressure, so we can't get it to cut without leaning on the plane uncomfortably hard.

Does this make sense?

Miles

Re: And a little more....

woodburnbob

>Miles, I appreciate your thoughtful information and ideas.

Steve suggests that the thread test may not really reflect the planing behaviour of a worn edge...

If Steve prepares a fresh edge that is maximally sharp (by his standards) with a 45 degree relief, his thread test will show a maximally sharp rating. Putting the same blade in a bevel-down plane having a 45 degree frog should show the blade to be maximally dull ...given the conditions. This, I think, is the extreme example of what you describe. Same included angle, different tool paths into two different objects to be cut. I don't see any disputes here.

However, the beauty of something like the thread test as an indicator of just sharpness (whether initial sharpenss or after dulling) is that it removes extraneous and circumstancial factors such as included angle and relief angle of the arris...within some practical range of included angles (e.g. 1 degree to 70 degrees) and trajectories of "presentation" (tool path) off perpendicular to the object (e.g. 30 degrees either way). My thinking is that within some range like this, severing the thread probably reflects sharpness independent of included angles, relief angles and tool path. Sheer spectulation on my part.

Granted the thread test won't say anything about performace of that edge once it is given a set of specific angles, tool paths, orientations, forces and materials to cut...and then dulled in one way or another.

My problem with defining dullness or loss of sharpness as only or predominately the emergence and widening of a zero degree relief is this: If I use a plane on white oak for 30 minutes the blade will be pretty dull. I can take it out of the plane and look straight down on the edge and reflect light from a nearby light bulb straight up into my eyeball. If I roll the blade either way from vertical to horizontal the width of that "candle" gives me some sense of the geometry of that dull edge. Certainly I have to remove the candle from all aspects of the edge if I hope to resharpen the blade. In spite of the photos from Steve and others, my impression of dulling is that an arris becomes a radius ... or in the most abstract sense, an infintesimal radius becomes larger. Maybe it's not an exact radius. Maybe it's more elliptical. Still, you see my observations.