Testing early stage edge retention

[Dog of War]

Like most of you I imagine, I've noticed that a fine edge quickly looses that high degree of sharpness pretty quickly. Cliff's recent thread comparing a Zubeng utility hunter and Caffrey MEUK got me thinking about this in earnest, so I decided to run tests on some knives of my own to hopefully study early stage edge retention.

My test method was pretty simple: starting with a freshly sharpened blade I would measure loss of sharpness from cutting cardboard by testing sharpness on newsprint. "Sharpness" here means the maximum distance from point of hold where the blade is able to push cut the paper (the higher the better.) My hope was that using newsprint as a test medium would offer good "resolution" on sharpness loss in its earlier stages. I also relied greatly of the reliability of the newsprint test method which I've been using a lot lately and only repeated tests or rejected results if I felt technique was poor or it seemed too much of an outlier .... meaning this is informal testing mostly to satisfy my own curiosity. To improve control while cutting and better limit the section of edge used I cut with handle lower than tip, giving a combined push cut and slice action, and with the blade also held about 45 degrees (left-right) to the plane of the cardboard to reduce wedging.

I'll jump right away to a graph and table of results so nobody has to wade through a bunch of details if it doesn't look interesting:

Table of data plotted above:

Following are descriptions of the knives tested with manufacturers' specs, edge geometry and brief comments:

Frosts traditional Mora (Mora #1) - carbon steel laminate ~61 HRC, 10 deg./side, 15 deg. microbevel. This knife was one of the more awkward during use, probably due to handle ergonomics.

Frosts 906 (Mora #2)- carbon steel laminate ~61 HRC, 10 deg./side, 15 deg. microbevel. Same steel as Mora #1, added to testing queue to compare after somewhat erratic findings with Mora #1.

Eriksson 945 (Mora #3) - 12C27M ~58 HRC, 10 deg./side, 15 deg. microbevel. This knife is a sleeper IMO, not only in edge retention but excellent handle ergonomics. Edge condition after cutting looked exceptional under magnification, best of the knives tested. This blade gave quite a few sharpness readings significantly higher than those reported, so results shown for this knife may actually be conservative.

Cara Cara - 8Cr13MoV ~61 HRC, factory edge about 20 deg./side, 22 deg. microbevel. Basically the factory edge with microbevel added to improve consistency a bit. Sharpness testing a bit more challenging due to coarser edge finish, but all things considered it looks like this steel with its greater hardness may compare well with some premium blade steels.

Calypso Jr. - VG-10 ~59 HRC, 12 deg./side primary, 17 deg. microbevel. From a user standpoint this is the easiest knife to use for cutting cardboard .... control was so good it may be that results are skewed a bit low due to ease with which cutting could be isolated to small area of the edge. Of the three knives showing any micro-chipping this was the most extensive but still only very, very slight.

BRKT Highland - A2 ~59 HRC, convexed with apex 10 deg./side and 15 deg. microbevel added.

Ka-Bar Dozier - AUS-8A ~57 HRC, 10 deg./side, 17 deg. microbevel. Selected as kind of a minimum standard for blade steel favored by many who post to BFC. High starting sharpness is easily attained with fine ceramic, edge condition after cutting looked excellent with no micro-chipping indicating performance was due to lower hardness and faster wear than others.

Becker Necker (BK11) - 0170-6C ~58 HRC, 10 deg./side, 15 deg. microbevel. Even having used this knife quite a bit lately and liking it a great deal I was surprised by its performance. Like the Eriksson Mora I rejected quite a few higher sharpness readings during testing.

EKA compact - 12C27 ~59 HRC, 10 deg./side primary, 15 deg. microbevel. Surprisingly low performance, especially considering results from the 12C27M Eriksson.​

There's a lot I see in these results that would be interesting discuss, since this post is already kind of long I'll mostly let others take it wherever they'd like to go (if anywhere.)

One thing important however, I think it's useful to judge the knives in the test by the average slope of the lines in the graph and not just end sharpness - slope reflecting the rate of loss of sharpness. This not only gives some correction for the different initial levels of sharpness, but just looking at the chart the slopes of the lines seems to invite extrapolation as to what would happen if additional cutting was done.

Ranking performace by the slope of the line doesn't change things a lot - sheds maybe a little light on relative performance between the three Moras, and suggests the Cara Cara may compare pretty favorably against both the VG-10 Caly Jr. and the A2 BRKT Highland:

Of course you could look at slope using just a portion of the data different section, say from the 35" to 140" mark, but absent more rounds of testing I think I'm already pushing limit a bit on what can be reasonably concluded from this.

One thing that does seem pretty clear, those of us who like a very fine polished edge for things like wood carving have to accept that more abrasive materials will quickly degrade that fine edge, so having knives dedicated to certain kinds of work may make a great deal of practical sense. IMO there doesn't seem to be much point in putting a really fine edge on a knife that's going to get used on cardboard or other abrasive materials. With a little practice using a Sharpmaker or similar it's not hard at all to get a decent blade to push cut newsprint at around 1.8 - 2" and that's probably more than adequate for an all-around utility blade .... in fact I believe with some blades raising a higher level of sharpness can lead to edge stability problems, resulting in an edge that in early stages becomes more dull than if it had been left not so finely sharpened.

------------

12/1/06 - Additional testing of the EKA at a slightly more obtuse geometry of 12 deg/side primary and a 17 deg. microbevel has giving much improved performance. Starting with a very fine edge that would push cut newprint at 4.05" from point of hold, the EKA was still push cutting newsprint at 3.25" after cutting 140" of cardboard (same stock used in all tests.) It's uncertain at this point whether there was a problem with the way the blade had been sharpened during the original test or if the more acute geometry in the first round left the edge unstable.

 

[Cliff Stamp]

First off all, an impressive amount of work and analysis, I would hope this is not a singular event. Secondly, the comments made about suggested changes to method are just that, suggestions, not mandates.

To improve control while cutting and better limit the section of edge used I cut with handle lower than tip, giving a combined push cut and slice action, and with the blade also held about 45 degrees (left-right) to the plane of the cardboard to reduce wedging.

So there was no draw and the cutting was constrained to a section of edge basically the width of the cardboard? Are the results presented of one run or an average of several?

... graph

The colors are a great help, in general with experimental data you want to see the measured points in addition to any models or smoothing curves. For the points themselves it is of extreme benefit to have a measure of the confidence of the data, usually the standard deviation in the mean or a scaled IQR for median data. Without this measure of uncertainty it is very difficult to tell if results are significantly different.

This blade gave quite a few sharpness readings significantly higher than those reported, so results shown for this knife may actually be conservative.

What exactly do you mean here, how were the measured values determined - a selective mean? A median is ideally suited for samples that are prone to outliers (low frequency but highly deviating) as it isn't sensitive to thier magnitude. You can use the standard deviation as a high end estimate of its uncertainty but it is better to use a scaled IQR. The formula's are easily used in any spreadsheet or decent calculator, freeware versions of each exist.

One thing important however, I think it's useful to judge the knives in the test by the average slope of the lines in the graph and not just end sharpness - slope reflecting the rate of loss of sharpness.

You can model the data by a simple power law which is easy to fit on freeware programs. It has been several years since I used spreadsheets but I recall that they could do it as well. This will give you both the estimates for the initial sharpness and the type and rate of degredation which use the entire data set to provide estimates for each attribute, this makes the predictions far more robust.

... absent more rounds of testing I think I'm already pushing limit a bit on what can be reasonably concluded from this.

Yeah, based on what I have seen I would want about 15 points to get a decent estimate of the relative behavior and that takes a fair amount of time especially with the volume of knives you are using. However as long as you keep the cardboard pool consistent and do them within a sensible time frame you don't need to do them all at once anyway.

.... in fact I believe with some blades raising a higher level of sharpness can lead to edge stability problems, resulting in an edge that in early stages becomes more dull than if it had been left not so finely sharpened.

If you are going to hone an edge so the bevels meet, then making it as free of defects and debris (make it sharper) would only make it stronger. However some cutting tools like cold chisels have radiused edges because if they were fully sharpened they would indeed be too weak. The sharpness for such tools is essentially zero compared to a knife edge. In general it is also the case that the optimal edge angle for edge retention isn't simply the most acute one possible because at some point it will just fold and thus a micro-bevel of the right angle will give optimal cutting lifetimes. However I would not call adjusting the angle effecting the sharpness, confusing the two is one of the main sources of misinformation and confusion about knives.

-Cliff

 

[sodak]

Very good work, DOW! Very interesting! I'm impressed with how long the edges held. How exactly did you sharpen them, sharpmaker, freehand, etc.? How fine were the edges/how fine was the grit?

How high is the bevel on the BK11? I've got mine at 15 per side, and it's about 1/4 to 3/8 inch wide. You're running it pretty thin, have you cut anything besides cardboard with it?

 

[Dog of War]

So there was no draw and the cutting was constrained to a section of edge basically the width of the cardboard? Are the results presented of one run or an average of several?

Right, while some slight draw or drift of the blade was unavoidable at times, holding the blade at an angle like I did so that a bevel cut was made, pretty much assured the central area of the portion of edge being used was always in contact with the cardboard.

Results are only one run, in fact I was surprised how much work it is cutting about 100 feet of cardboard while trying to control the knife as much as you possibly can.

The colors are a great help, in general with experimental data you want to see the measured points in addition to any models or smoothing curves. For the points themselves it is of extreme benefit to have a measure of the confidence of the data, usually the standard deviation in the mean or a scaled IQR for median data. Without this measure of uncertainty it is very difficult to tell if results are significantly different.

This was my first use of OpenOffice for graphing and haven't figured out yet how to add marks to the points themselves. Obviously what I've presented here is limited, not rigorous enough that anyone should take the results as gospel. And these of course are just single examples of each knife. But I think it at least indicates the approach is viable and by gathering more data and adding improvements like you suggest could produce reliable, valid results.

Still testing like this, even if limited and done sort of casually, is a quantum leap for the knife user like myself. I feel I've learned more about edge retention and related blade performance in the last few months than I'd learned in all the time before. Just the effort to be systematic and quantified forces a lot of subjectivity out the window. Plus the feedback you get doing the cutting and examining edges with a microscope gives information that's bound to be useful for understanding results and later trying to optimize individual blades and better understand potential limits in their scope of use.

What exactly do you mean here, how were the measured values determined - a selective mean?

This is where my method I think could be seriously criticized. Push cutting newsprint I believe is a pretty sound method of testing sharpness, it's even surprising how often you get the same result repeatedly, +/-.05". But there are exceptions, sometimes the newsprint will be less stiff from handling, and I've noticed that heavy inking on both sides affects results.

Since I always start the sharpness measurement trying to cut the newsprint further out on the paper, sometimes you get a higher cut that's probably due to a weak spot in the paper or an irregularity in the blade that's extremely sharp. So if I couldn't also get the blade to push cut ~.10 and ~.20" closer to point of hold I rejected it outright. I would then either trim a clean edge on the newsprint or take another piece from the same page and repeat the test until results were consistent. Instead of averaging, I used the most conservative data reflecting the edge's repeatable push cutting ability.

It was actually two of the better performers, the Becker and the stainless Eriksson, that made me realize in the middle of testing the shortcomings of doing it this way, these two knives had the most unexplainable high sharpness readings and so probably wound up getting sandbagged on the results somewhat.

Yeah, based on what I have seen I would want about 15 points to get a decent estimate of the relative behavior and that takes a fair amount of time especially with the volume of knives you are using. However as long as you keep the cardboard pool consistent and do them within a sensible time frame you don't need to do them all at once anyway.

Not really knowing what to expect I didn't define the test too well before starting. At this stage testing a larger number of knives didn't work out so bad - basically IMO results show six very good blades with performance clustered pretty close together, and two blades that, though very serviceable for many purposes, aren't in the same class.

A much better approach would be to start by limiting the number of knives being compared to just two or three, then doing more runs and using a defined and consistent way of averaging data.

Very good work, DOW! Very interesting! I'm impressed with how long the edges held. How exactly did you sharpen them, sharpmaker, freehand, etc.? How fine were the edges/how fine was the grit?

Thanks, Sodak, I can't tell you how interesting this was for me. Some results I kind of expected but probably more surprised than anything by how some of the knives fared. The ability of the better knives here to hold a useful edge really is much better than thumb and thumbnail tests lead you to believe. Particularly you look at the knives where the slope of the line levels out in the later rounds and then figure this is just one small section of the blade, these knives obviously can do a lot more work than a lot of us will ever do between sharpenings.

Except for the Cara Cara the primary bevels were done with a 240 grit AO benchstone held 10 or 12 degrees off vertical in a contraption I made that works like a heavy duty one sided Sharpmaker. Microbevels were done with Crock Sticks which I assume are around 1200 grit. No stropping was done, a couple knives used that had been stropped previously were "detuned" for testing. The Cara Cara had the coarser factory edge with a slight microbevel added. The forward part of the BK11's edge had previously been convexed with 3/0 emery paper but that part of the edge wasn't used in testing.

How high is the bevel on the BK11? I've got mine at 15 per side, and it's about 1/4 to 3/8 inch wide. You're running it pretty thin, have you cut anything besides cardboard with it?

Pic of the BK11:

I've had it at this geometry, 10 deg./side and 15 deg. microbevel for maybe a month. From the tip back about 1.25" has been lightly convexed, mostly because I found the 10 deg. edge practically matched the blade grind near the tip and did this just to clean it up.

It's seen EDC use and cutting of some pretty tough wood, ash and branches from an ornamental plum/cherry, and it's been fine. Before that I had it at 12 degs. plus 17 deg microbevel and I know that is very durable. There was just the tiniest bit of micro-chipping seen after the cardboard tests so I'm going back to the 17 deg. microbevel for general purpose use since a big reason why I carry the Becker is it's robustness for a knife this size.

 

[Cliff Stamp]

.... while trying to control the knife as much as you possibly can.

Any variation in method will of course cause a corrosponing variation in the end results but because you are essentially averaging over 100 ft of cardboard the mean cut angle/speed is pretty much exact. Through 100 cuts the variation in the mean is actually 10 times less than through 1 cut. Lets say that from one cut to the next you drift by about 10 degrees, you can expect the mean angle to be within 1 degree. The big differences are the systematic effects like the initial sharpness, the carbide structure of the edge, a high inclusion spot in the cardboard, etc. .

But I think it at least indicates the approach is viable ...

Very much so. One thing which has to be kept in mind is that any attempt to be objective makes a huge improvement and that any further steps are always just a refinement. It is similar to comparing how productive you are with just a $2 knife vs your fingers vs how much more capable you become with better knives.

The main conflict is a wider data set or a smaller but more precise/accurate one. For someone starting off you always go wider for a number of reasons. Consider spending a year comparing ATS-34 vs VG-10 in detail and it turns out that neither of those have the properties you want and in a short period of time neither of them are used any more.

Instead of averaging, I used the most conservative data reflecting the edge's repeatable push cutting ability.

That is averaging, there are many ways to average data. The most common one, the arithmetic mean is the value which gives the least squares residuals. It is actually a very poor way to average physical data because it tends to be prone to large outliers. What you are doing is essentially taking a skewed median which is perfectly fine as long as you define it is what you are doing and ideally show a sample data set and the resulting average.

A much better approach would be to start by limiting the number of knives being compared to just two or three, then doing more runs and using a defined and consistent way of averaging data.

All initial lab work is a broad and rough experiment which you use to firgure out how to actually do what needs to be done. As you learn more you can refine your methods accordingly. Pretty much the largest priority is to repeat. How often you choose to do so is basically up to you but the gains are incremental; two data sets (30% reduction in variation), three (43%), four (50%), five (55%), etc. . So basically three is a practical limit and you don't gain much beyond that unless you are willing to go very large and quite frankly I think you would be better off looking at more knives. On that note it would also be better to do one run on 5 Mora knives than 5 runs on one. The former average would be far more meaningful than the latter.

It's seen EDC use and cutting of some pretty tough wood, ash and branches from an ornamental plum/cherry, and it's been fine.

Slicing or chopping/splitting.

-Cliff

 

[Dog of War]

Really appreciate your comments, Cliff, IMO great insight for anyone thinking about "upping their game" in evaluating blade performance .... or pretty much anything else for that matter. I guess some people think you're too demanding but I'm often amazed by your ability to find the value in whatever information is presented.

One thing which has to be kept in mind is that any attempt to be objective makes a huge improvement and that any further steps are always just a refinement.

This work really drove this point home for me. Several knives tested had been used casually side-by-side before, and I had formed some rough opinions about edge retention based on that. Not that I totally distrusted those observations or felt I was being biased about it, but since nothing was measured in a way that could be recorded no future comparison is possible. So when you do casual work like that it quickly becomes little more than an opinion or belief you can't quantify or expand on, or even explain or justify to anyone including yourself.

Because of the effort to be objective, I was just as surprised by results I might have expected as I was with those that I didn't. The value of this kind of objective, quantified testing is so obvious once you start using it, you wind up with a great deal more confidence in what you've done and any conclusions drawn.

It's [BK11] seen EDC use and cutting of some pretty tough wood, ash and branches from an ornamental plum/cherry, and it's been fine.

Slicing or chopping/splitting.

For me an EDC knife doesn't tend to see much slicing use or chopping/splitting, but it's pretty common that I do heavy push cutting on cables, wood, some plastics. Of course there's the usual scraping, cutting tape, etc. over the course of the day. Based on edge damage the most demanding thing done with the Becker in recent weeks was cutting 14 ga. electrical wire, and that only left a couple small impacts of about .00005" or less with the 10 deg./side geometry.

A habit of mine is to do test cutting with new, reprofiled and freshly sharpened knives on clean wood. This includes quite a bit of push cutting, and what I guess I'd just call whittling - stuff commonly seen in bushcraft type work. I consider this a kind of minimum standard for a blade, any problem with wire edge or tendency to burr is revealed quickly as are other gross problems. The ash and ornamental are hard enough that not a lot of work is needed before edge degradation detectable by hair shaving, thumb and nail tests.

 

[Cliff Stamp]

I guess some people think you're too demanding ...

I am pretty demanding of myself but that is just my background. Of others and the work they care to present, all I would demand is the truth. Whatever they care to do, in whatever way they care to would be welcomed as long as it isn't a biased attempt to hype or condemn a product.

The value of this kind of objective, quantified testing is so obvious once you start using it, you wind up with a great deal more confidence in what you've done and any conclusions drawn.

Yeah, the downside is the work. One thing to note is that if you are not in a rush to collect the data then you can obtain the results with far less effort. Just sharpen the knives and measure the results at the end of each day. Don't make any effort to constrain the work you do, it is not necessary. Just keep a log book where you record the sharpness daily and when you need to resharpen and start over or cycle in a new knife.

After awhile the averages (which are trivial to calculate automatically in a spreadsheet) will stabilize and you now have a very large data set and all you did extra was to just note the sharpness at the end of the day for one knife, maybe a minute or so to record and log the data. Basically you are sampling with a large variance so you need a large number of samples to get a fairly precise result. However I would expect that even in a few weeks patterns would emerge clearly and in months they would be very clear.

-Cliff