Edge retention on carpet : cKc-X, O1, S30V, k390

[Cliff Stamp]

I recently was helping a friend who noted they were throwing out a pile of used carpet and like any sensible person I saw an opportunity to cut it up into little tiny strips while listening to Skynard :



Knives :



-Kyley Harris fixed blade, cKc-X (steel not released)
-Ozark Tradition Knifeworks EDC, k390 / 64 HRC
-Spyderco Paramilitary, S30V
-Uddopukko/Enzo, O1

The method :

-all knives zero ground to a very light convex, edge bevel starts to increase curvature at ~0.010"/6.5 dps
-15 dps micro bevel, 600 DMT
-carpet cut with a 2" section of blade, cuts were 2" deep
-cuts were made through random selection on the same section of carpet
-blinding was used (the individual measuring the sharpness didn't know how many cuts were made)
-sharpness was measured in multiple different ways (35 g/Bergia, 500 g/heavy jute, 2" slice on 3/8" sisal, visual inspection, etc.)
-126 slices were made, four rounds of each to produce average / total values
-a Naniwa Superstone 400 was used to reset the edge, pressure on the stone was 0.4 to 0.8 psi

The results :

-final sharpness was ~1.5% of optimal
-cuts on 3/8" sisal at this point were ~10/11 lbs
-all edges showed a reflection of light (the amount decreased in the order of the list below)
-the edge retention was much closer than the resharpening times (# passes)

In some detail :

• cKc-X : TCE :0.27 (2), PPS (total) : 550
• Uddo/O1 : TCE 0.36 (6), PPS (total) : 1115
• Paramilitary/S30V TCE (1.5%) : 0.48 (5), PPS (total) : 2025
• OTK/k390 : TCE 0.32 (2), PPS (total) : 4250

The PPS count is the number of passes needed on the waterstone, it is the total of four runs. The TCE is a measure of edge retention, it is calculated as :

sum = amount of material cut (in meters) * sharpness (as a percentage)

It basically is a weight average of the sharpness during the cutting, the higher the TCE the stronger the edge retention. In this case the final sharpness was very low, ~1.5% thus the TCE is for extreme cutting, this would be like cutting a km of cardboard and thousands of cuts of hemp/sisal.

Comments :

The results are fairly straightforward aside from the k390. It had an acid treatment to force a paina (from the maker), I think I am still seeing some effects of that. I have some more carpet left and I am considering :

-doing a few more runs to get more precise values
-do a super long runs with a couple of blades to extreme low sharpness (<0.5%)
-check a couple of more UHC steels
-try a different angle/edge finish (25 dps/mxf DMT)

The last three are more interesting as I think they will show different relative performance ratios.

[The Mastiff]

-final sharpness was ~1.5% of optimal

I'm struggling to follow Cliff. What is Optimal? Is it different for each knife/steel or the same. How is it determined?

Thanks,

Joe

[Cliff Stamp]

Update :

I added another four runs with the Farid / 121REX Mule which I zero ground to the same profile as the others. Results :

• cKc-X : TCE :0.27 (2), PPS (total) : 550
• Uddo/O1 : TCE 0.36 (6), PPS (total) : 1115
• Paramilitary/S30V TCE (1.5%) : 0.48 (5), PPS (total) : 2025
• OTK/k390 : TCE 0.32 (2), PPS (total) : 4250
• Farid/121REX : TCE 0.31 (2), PPS (total) : 2000

The 121REX blade is kind of curious because even though it has very low grindability the edge takes very little macro level damage and thus there is very little material needed to be removed to sharpen and thus the pass count isn't among the highest. I believe this is mainly due to the very high percentage and very high hardness martensite.

Graph :



Note the edges do not blunt in the same manner. The O1 :



-no chips
-medium rolls (relative to the other blades)

The S30V :



-less rolling
-very small chipping

The 121REX :



-almost no rolling
-heavier chipping

I think a 25 dps/6 micron run is in order to see what happens to the relative performance if the angle/polish is high enough to stabilize the higher carbide steels.

[Cliff Stamp]

Here is what happens if you scale the edge retention by the amount of grinding necessary to apex the blade :

[Mike157]

Cliff, thank heavens you're "sensible" enough to see the "value" in used carpet and perform these tests. My only concern would be in not getting too carried away with something like Free Bird and trimming fingers along with carpet.

So if I understand this correctly, your results show that the S30V blade had the highest edge retention (0.48), so for this type of cutting would be expected to cut the most carpet before reaching the prescribed low level of dullness. But given the amount of work to bring back the edge, the cKc-x might be the one to use if someone were actually cutting similar materials for a living and time and money were involved (given the significantly fewer passes needed to bring the edge back).

Please correct if I'm misunderstanding. Thank you. Mike

Not to hijack the thread, but based on your earlier thread about thin blades and the Robert Herder paring knife.... given a Sharpmaker 15 dips bevel per your suggestion, it has become one of my favorite apple slicing/eating blades along with my Stretches in blue and ZDP 189 steels.

[Cliff Stamp]

My only concern would be in not getting too carried away with something like Free Bird and trimming fingers along with carpet.

I am not religious, but one does not work during Free Bird.

So if I understand this correctly, your results show that the S30V blade had the highest edge retention (0.48), so for this type of cutting would be expected to cut the most carpet before reaching the prescribed low level of dullness. But given the amount of work to bring back the edge, the cKc-x might be the one to use if someone were actually cutting similar materials for a living and time and money were involved (given the significantly fewer passes needed to bring the edge back).

Yes, I believe the angle/edge finish is very critical though. I am going to check that shortly as I believe I can move that sweet spot that forms in edge retention up/down that carbide scale by adjusting the angle/grit finish.

... and the Robert Herder paring knife.... given a Sharpmaker 15 dips bevel per your suggestion, it has become one of my favorite apple slicing/eating blades along with my Stretches in blue and ZDP 189 steels.

Yeah it is something else, Roman actually makes large chopping blades out of S5-type steels which are not a whole lot thicker ground.

[me2]

-final sharpness was ~1.5% of optimal

I'm struggling to follow Cliff. What is Optimal? Is it different for each knife/steel or the same. How is it determined?

Thanks,

Joe

Optimal sharpness is as sharp as you (the tester, Cliff in this case) think you can get it. For these knives, it is likely somewhere near push cutting thread on a scale at 80-100 grams, or slicing light thread in about 0.5 mm of edge length (for aggressive edges). The knives being tested need to be near the same sharpness or that will provide an advantage in the early stages of cutting. For instance, I usually test for push cutting sharpness. The highest I can regularly get all but a handful of my knives (which have known issues) is push cutting thread on a scale at 17 grams (different thread from Cliff's). So, I would set optimal sharpness at say 14-15; just out of reach, and low enough to not have anything go off scale.

[Mallus]

Just to make sure (and I think I've asked this before, but couldn't find the answer): was optimal sharpness in principle the same for all bladesteels or was it steel spesific? I.e. Is the starting point 100 % or variable?

[Cliff Stamp]

So, I would set optimal sharpness at say 14-15; just out of reach, and low enough to not have anything go off scale.

Yes exactly. In the above I am using the length of light cord cut under a specific draw but the principles are the same. Note that where you set optimal sharpness doesn't change any of the relative performance it just makes the sharpness easier to interpret and semi-standard. For example at some point I will run out of the cord I am using and have to change. I have already calibrated another cord so I know what measurement on it gets the same one as the Bergia so even though the numbers are different the sharpness level is the same.

Note it is also determined relative to the grit, in this case (the carpet cutting) I was using a 600 DMT.

For just a loose/quick reference as to what 100% means or looks like :

http://www.youtube.com/watch?v=yoLV30_ZfQg" target="_blank

The sharpness demonstrated there is about 10-20% of what the 600 DMT finish is at optimal. It is possible to get it to match the DMT but you have to be very careful as it is a soft and muddy stone and any slurry will reduce the sharpness so if you really want to get it a lot sharper than what is in the video you have to flush almost after every pass.

[Cliff Stamp]

... was optimal sharpness in principle the same for all bladesteels or was it steel spesific?

All steels are scaled by the same factor, the initial sharpness is within the random spread among the steels. It takes more work to get the steels there in terms of number of passes and the detail needed on burr removal on the 600 DMT and in general some of them take much more care in the shaping. For example :

• cKc-X : very easy to grind no concern in shaping, a little concern about burr through high ductility in forming
• O1 : easy to grind, no concern in shaping, no concerns in forming the apex - just very easy to sharpen
• S30V : hard enough to grind that particular stones are needed, apex forms well on DMT
• k390 : very hard to grind, stones for shaping are particular, apex takes more passes on DMT, no concern about burr minimization
• 121REX : stones for shaping are very particular, apex takes even more passes on the DMT, no concern about burr minimization

In some detail, as you move into the very hard and high carbide steels it is easy for the apex to start to break apart with the shaping grit and if you don't take care here you will see the fracture have an influence in lowing the edge retention/durability even if you micro-bevel because the fracture is deeper than what you see. For example these are some shots of MaxaMet which is similar to 121REX being an ultra-high carbide steel :



This is the edge apexed with the Bester 700, note how the apex breaks apart larger than the scratch lines. If you go lower than the Bester 700 the apex can literally explode. Here is what happens when I apexed it with a TASK which has a grit rating of ~100 :



This however is what the apex should look like before the micro-bevel to get maximum performance, it was produced by a Naniwa Aotoshi :



But that is a 2k stone and very slow cutting. The Naniwa Superstone 400 is a much more coarse stone but gets you to a very nice apex level and cuts fast enough it can be used for shaping :



Now if you have the time/inclination, a more optimal approach is to work with say the Naniwa Superstone 400 until you just start to see the light be removed from the apex and then switch to the Aotoshi. But the problem is that the very slow cutting action means if you don't judge the transition exactly right you will be grinding a long time and doing little to nothing. This is why most people just maximum burr sharpen as it is very fast compared to trying to minimize burr sharpen.

But you waste a lot of steel if you sharpen that way (constantly create heavy burrs) and when I really start to work with a knife I can sharpen it several times in one day, if I sharpened using heavy burr formation I would increase the rate of steel loss many times to one and burn up the knives much quicker. The way I sharpen doesn't remove any more steel than what was lost in the cutting itself, speaking of the width of the knife of course.

[The Mastiff]

Thanks ME2 and Cliff. Was wondering how that worked out and how one would find optimal for the different steels.

Joe

[Mallus]

Thanks Cliff, It seems it's rather easy to get sub optimum performance out of the very high carbide steels as you'll have to be very carefull in shaping the apex not to stress it in the process. Diamond abrasives would seem to help, but there the presense of random high points (in my coarse DMT stones at least) have their own ill effect in the form of deep, weakening scratches that would need to be negotiated somehow.

The more I sharpen, the more I appreciate the ease of sharpening.

[Cliff Stamp]

Diamond abrasives would seem to help, but there the presense of random high points (in my coarse DMT stones at least) have their own ill effect in the form of deep, weakening scratches that would need to be negotiated somehow.

I used to think that because the diamonds will easily cut the steel and carbides, however after looking at the apex carefully and just comparing a bunch of them, diamond abrasives are very nice for finishing an apex to a specific grit/pattern however they are fairly problematic to shape as they produce very harsh scratches which can induce a lot of fracture. This is a common complaint among the straight razor crowd about diamond abrasives. As they are working with very low angles in general the problem is magnified with them.

The more I sharpen, the more I appreciate the ease of sharpening.

Kyley Harris has made the argument that for practical purposes the kind of comparison I am doing is biased in favor of high carbide steels because I make sure the initial sharpness is similar with all but this means far more time has to be spent on the high carbide steels. If you did a comparison where you sharpened to a specific time instead then the results would shift in favor of the lower carbide steels. That approach he is talking about is not uncommon in wood working which is one of the reasons why the high carbide steels have made very little ingress there because the time to sharpen is critical and you have to be able to obtain and hold a very high sharpness as otherwise it degrades the finish of the wood.

[avocadobbq]

Quick question: is ckc-x stainless?

[Cliff Stamp]

Quick question: is ckc-x stainless?

No, I have done basic corrosion checks on the forum. It behaves like a mid Cr steel.

[Chum]

-cuts were made through random selection on the same section of carpet.

Could you explain this a bit more por favor? So you take a random selection of carpet and make an equivalent amount of cuts on that section with each knife?

You have shown before that just a little bit of dirt can greatly effect edge retention tests. Carpet has a habit of getting dirtier in random areas, and it isn't always visible. How does your testing remove this potential variable? Do you dump the highest and lowest results on each batch of testing, for instance?

[Cliff Stamp]

Could you explain this a bit more por favor?

I have a few types of carpet, some are much heavier than others. I picked a specific type when I started the trial and cut it up into ten equal sections. When cutting had to be done each interval, a random number was generated from 1-10 and that determined which section was used for the cutting in that interval. There were 8 intervals in each round (which could be randomized as well in terms of length - how many pieces were cut).

As the carpet had a very high blunting rate, it could easily be cut until the sharpness was less than 1.5%. This essentially means the blade could not make a cut on the light cord with a full 2" draw. At this point it is fairly dull,it will slip on a tomato, meats/fats, and requires much more force (or draw) to cut ropes/fabrics etc. . You can also see the edge at this stage, it reflects light and shows deformation/chipping depending on the steel.

To get this kind of blunting on cardboard it takes on the order of a km of it, on ropes it also is a LOT as in thousands of cuts. But with used materials the amount you can cut is severely reduced because of the dirt.

How does your testing remove this potential variable?

That strictly isn't a variable, the variables in an experiment are the things you are changing and the things affected by it. In this case the edge retention is one variable (dependent variable) and the length of carpet cut is the other (independent variable). What you are talking about is a source of error and they can never be removed, all you can do is try to prevent them from being systematic (that introducing a bias) and minimize the scatter through averaging or similar techniques.

For example if I didn't randomize the cutting it could be possible (even likely) that one knife was used to cut through a very dirty section and another knife a very clean section. This could mean that the result was not indicating the properties of the steel but the carpet being cut, i.e., it was biased. However if you do it the way I did with 4 trials of 126 cuts each with random selection on multiple intervals, the probability of a systematic bias is very low. If this isn't obvious then consider this simple experiment :

-take a ten sided die
-roll it ten times
-add up the numbers
-now repeat that ten times
-average the results

If you did this and I did it would you expect the numbers to be closer together than if we did this :

-take a ten sided die
-roll it once
-multiply the result by 10

The way I do the cutting is like the previous one with the multiple rolling and averaging. If you think about it a little you can see the more times you roll the dice the closer together the results get. This is extremely easy to simulate in a spreadsheet program, if I do this in calc for example I get the following averages for two runs :

First run : 59

Second run : 56

(the expected mean result is 55 so the results will scatter around that)

Note how close the averages are even though they come from adding up a dice which can produce a result from 1 to 10. This is a huge scatter in the individual results but the net or total you are measuing shows very little random scatter. In regards to cutting this means you can cut materials which produces very high random blunting results but yet the final result is very consistent from run to run. Note that even that simple setup I described took a variability of 1000% and reduced it to less than 10% spread.

That is the power of science, it doesn't afraid of random errors. It laughs at the random errors, in fact it uses random spreads all the time. However it greatly fears the systematic error though it has weapons to use against it as well such as a +2 mace of double blinding, and an angelic two handed sword of peer review.

--

For those so interested I don't do simple averaging because even if you do interval randomization you can still get occasional strong outliers. You therefore need to have a way of getting central tendancy which is much more robust. I use either median based statistics or a weighted average where the weights are usually based on the absolute deviations. This kind of measurement is extremely robust meaning if you repeat the experiment multiple times then you consistently get the same results. In general I will favor statistics which increase accuracy at the cost of precision, i.e. I will minimize bias even if it increases scatter.

[Chum]

For example if I didn't randomize the cutting it could be possible (even likely) that one knife was used to cut through a very dirty section and another knife a very clean section.

The thing about used carpet is how much random material it can retain. For instance, if you were to cut random samples of our shop carpet you are likely to patches of dirt, patches of metal filings, patches of various adhesives, areas worn by various chemicals, oily patches... pretty much every kind of filth you can think of.

If you were to take 1'x1' section, randomly or otherwise, and cut up that piece with one knife, then move onto a new 1'x1' section and cut that with another knife... I'm not sure you would ever get two samples with similar edge wearing characteristics.

[Cliff Stamp]

[...]

If you were to take 1'x1' section, randomly or otherwise, and cut up that piece with one knife, then move onto a new 1'x1' section and cut that with another knife... I'm not sure you would ever get two samples with similar edge wearing characteristics.

Yes, that is why I don't do that because the results would be strongly systematically biased. I look at the cumulative effect of ~100 cuts and that cumulative effect is very stable over numerous runs. In cardboard it is similar but the cuts increase to ~1000 cuts.

Just do the experiment with dice, it should become apparent very quickly that even if you roll dice which has a huge spread (use a 20 sided dice for example) the sum of a large number of dice rolls is very stable.

[Chum]

Yes, that is why I don't do that because the results would be strongly systematically biased. I look at the cumulative effect of ~100 cuts and that cumulative effect is very stable over numerous runs. In cardboard it is similar but the cuts increase to ~1000 cuts.

I am interested in the exact procedure you take to make these tests. I apologize but I can't visualize it from the info you have provided.