Easy way to measure sharpness points.

[nozh2002]

This is awailable for everyone simple method to measure sharpness at single point of edge.

This does not require complicated equipment or math to calculate results, anybody can do this. Anybody can test blades.

What is needed:
1. Spring scales (I loan one from kitchen) I get used to gramms, but it does not matter.
2. Thick Cotton Tread (I have Caron Grandma's Best 100% mercerized cotton style 295).
3. Scotch tape.
4. Grid paper.
5. Pen.
6. Thread holder - I carve one from wood - only one thing you need to make youself.

What to do.
1. Create test-card - on grid paper write scale numbes vertically (in my case from 10g, 20g, ...200g).
2. Mark point on the edge with pen to test.
3. Stretch and tape thread over thread holder.

4. Put holder on scle and tune scale to show 0.

5. Carefully cut thread by matked edge point slowely pushing blade down and watching scale measures. Remember reading when thread was finaly cut.

6. Put birdy in test card at the row with the measured number.

7. Repeat step 3 to 6 at least 21 times.

8. From the top count 11th birdy on test card - it will be median (you may count from the bottom eather - will be same birdy). This is result of measurement - sharpness. (you may repeat measurment more then 21 then median will be half of this number+1)

Never cheat with single reading, if you are not comfortable with results for some reason - right thing to do increase number of measurement. If you check median is pretty stable with big numbers of tests - more tests better result.

Of course different people probably will have different absolute results, but at least anybody can compate two of his own steels and see which performs better. Or ake to brand new knives out of the box anc compare factory charpness etc...

Thanks, Vassili.

 

[MORIMOTOM]

i guess i am just a simpleton. if my knives cut magazine paper cleanly or shave hair off my arms, im happy.

i thought this was going to be a 'simple' thread. a little to involved for me.

but thx.

 

[nozh2002]

Example:
Kershaw Leek ZDP out of my pocket, resharpened once by me and being EDC for some time:

...
90 ++++++++
100 ++X+++
110 +
120 ++
130 ++++
...

Result is 100 - it is sharp but used a little bit.

Spyderco Military brand new, out of the box
...
50 +++++
60 +++++X++++
70 ++++++
...

Result is 60 - Very sharp!

ZT-302 never sharpened, used as big EDC from August (5 monthes)
...
140 +
150 +++++
160 +++
170 +Х+
180 +
190 ++
200 ++
210 ++
220
230 +
240 +
...
Result is 170 - not too sharp

Thanks, Vassili.

 

[Cliff Stamp]

Thick Cotton Tread

You want the thread to be as thin as possible, the greater the area of the thread the more you are measuring the influence of angle/thickess of the edge. You will also want to include a measure of the variance in your results, for medians you usually scale the IQR to estimate a standard deviation. Or you can just use the standard deviation which will be a conservative overestimate.

-Cliff

 

[nozh2002]

Example 2:

Military was sharp out of the box with factory edge and so I deside to test it.

I take 1/4"x4' piece of basswood - pretty soft wood with very even structure. cut it at 45 degree, so I cut it with mmilitary under 45 degree to fiber (length about 6") and measure sharpnes after one clean cut full length at the market point of the edge, and then after 5 same cuts, because edge losing sharpness not lineary.

This is my results.

it is very interesting to see that edge going little bit up and down. I noticed this before and now so I think it is may be due to the nature of loosing sharpnes. Hypothesis is may be initially tip of the edge bend and then breake off and this is why it gets up a bit each period.

Whatever it is, but for me this is indication that method really work and gives excellent measurement in numbers, which is way better then shaving sharp or paper cut sharp, especially if you like to see which steel is really better or how good this or that sharpenning technic etc. Cutting paper is just not precise enough to really answer this kind of questions. But this method shus up-and-down nature of sharpness decrease which was surprise for me, so this is really good instrument for any kind of blade, steel, sharpening testing.

Thanks, Vassili.

 

[nozh2002]

You want the thread to be as thin as possible, the greater the area of the thread the more you are measuring the influence of angle/thickess of the edge. You will also want to include a measure of the variance in your results, for medians you usually scale the IQR to estimate a standard deviation. Or you can just use the standard deviation which will be a conservative overestimate.

-Cliff

Point is that I want this to be as simple as Kalashnikov so enybody can do this in any household, without any math, calculators etc. Median is stabliest and simpliest solution - you do not need any math knowledge to understand it and use.

Again - anybody can do this not only mathematition, just fill test card and count 11th byrdy. I probably should not even mention that this is median, not to confuse people.

You see - I write it step by step, so it is very simple.

Thanks, Vassili.

 

[Cliff Stamp]

it is very interesting to see that edge going little bit up and down.

This is why you need variance measurements, what you are seeing in the above could simply be random deviations, and based on what I have seen in the past, with over six years of doing such sharpness measurements, the changes are well within the precision of such measurements.

...I want this to be as simple as Kalashnikov so enybody can do this in any household, without any math

Yes, however you can not simplify a process to a point it lacks meaning, and measurements with precision estimators are actually undefined. You can just count as well to calculate the IQR. If you take 10 measurements then just use the second and eight measurements and half the difference between them. Or if you want to do a more coarse estimate, assuming the distribution is symmetrical, which it is usually, just take the difference from the eight measurement to the fifth.

-Cliff

 

[Sal Glesser]

Very nice Vassili. Thanx for sharing.

sal

 

[Lavan]

Well, so much for my ongoing tests of the "Drop height to skin laceration ratio" method I was going to announce to the world after I healed.

 

[JSR]

I had my method of depth of cut on my hand/fingers weighted with the length of time for me to stop bleeding...but I'm starting to run out of fingers and hand area. I think it's time for me to adopt a new test method (tho I have a feeling I'll continue to occasionally use my old test method, most likely unintentionally).

 

[nozh2002]

This is why you need variance measurements, what you are seeing in the above could simply be random deviations, and based on what I have seen in the past, with over six years of doing such sharpness measurements, the changes are well within the precision of such measurements.

In this case, seeing this behaviour, I as I recommended increase number of measurement, which stabilize median more. Nature of this measurement allows me to do this - number of test not limited and so I can increase precision. I am not sure for 100% but I saw this many times before thinking exactely like you sad, but then I start thinking that there is not to much we know about mechanics of dulling so why it can not be sinusiodal.

Point is that I have enough data to came up with this hypothesis, which of course should be proven by numbers of tests.

Yes, however you can not simplify a process to a point it lacks meaning, and measurements with precision estimators are actually undefined. You can just count as well to calculate the IQR. If you take 10 measurements then just use the second and eight measurements and half the difference between them. Or if you want to do a more coarse estimate, assuming the distribution is symmetrical, which it is usually, just take the difference from the eight measurement to the fifth.-Cliff

Well, lets not talk in abstract terms, this is pretty detailed process I described, plus I gave enough examples to profe that it works in that cases as expected and plus it exposes (may be) some unkown nature of dulling - again I have enough data to announce this as a possible testing method, awailable for anybody. So if I claim it simple it does not mean automatically simplified to meaningless point - even paper cutting and shaving is not meaningless while they pretty simple. I better call it awailable for everybody then simple, there is some math behind it so it is simple only in terms of use.

I am mathematition, dealing with numbers and programmer. If you observe how numbers behave, without assumption, then you may notice some patterns. I tried different things - more then a year ago it was similar thread and I came to conclusion that median is best parameter which just came more and more precize with increasing number of tests - and this is what used in housing market to estimate average price.

Thread cutting results is similar - it is not simmetrical - sometimes you may see results very high (never very low) due to the random nature of edge dulling or shaking hands or thread being stronge in that poing or something and so only median is more stable - more tests done even one of them is too high. Again this is household sciense - cost of test itself does not matter, if real science it may cost millions to do additional measurement, but not in this case.

 

[Alberta Ed]

Very clever. I rely on the old shaving test mostly, and I know it's really sharp if I cut myself and don't notice anything until the blood starts flowing.

 

[Cliff Stamp]

Well, lets not talk in abstract terms ...

What I described isn't abstract, it is basic math, it is how measurements have to be compared when there are random influences which are of magnitude to the differences being argued to be significant.

If you observe how numbers behave, without assumption, then you may notice some patterns.

In order to attribute correlation to those patterns you need a measure of precision, and yes, without this the measurement is meaningless because it is impossible without that to compare results. This is first year statistics and fundamental to any measurement based science.

Now yes the trend is nonlinear, I noted that years ago and even showed the specific numerical model which fit the behavior and interpreted the variables of the equation in terms of physical properties of the edge. However the noise that you are seeing is just that, noise.

That being said, blunting does actually happen in the manner you describe, but not as rapid. I noted this as well and showed that you can for example see it in the very precise CATRA data. However in order to support that conclusion you need to be able to prove the deviations are larger than the random spread.

I noted this to you quite some time ago when you posted up similar data and noted the blade getting "sharper" and then "blunter" and I suggested then that if you had measurements of variance you would see that these are not real effects. Just do a monte carlo spread with a given variance and you will see these kinds of deviations.

The non-variance deviations tend to be much longer in duration and are cyclic. It is why you should generally do 3-5 trials and get the median on all of them to stablize the results. Anyone one one is subject to systematic deviations from such conditions as the sharpening, condition of the edge, stock material differences, etc. .
-Cliff

 

[nozh2002]

Well, I think as any hypothesis, this one should be proven or rejected by numbers of experiments, so I will perform eventually more of them and we shall see. The only way to see is experiment any arguments may have thousands contrarguments and so on. I hope that somebody else will try this method also (this is the point of all this) and then we may see more independent results.

So far I am convinced that method itself is working better then shaving and paper cutting. I will continue and will post more results here. Only what I hope is that sombody else will do the same, because it is so simple.

Thanks, Vassili.

 

[HoB]

I use the same setup, but I use rayon or mercerized thin cotton thread. It is very important, in particular for very fine edges that the thread is very thin. I have produced visible damage to the edge of a straight razor with the thin cotton thread. Both threads are very consistent. The problem with the rayon is that the thread easily unravels after the first strand is cut. The thread will then spread out and offer a much larger surface area resulting in erroneous results. If you make enough cuts, it is pretty easy to identify those though and discard them. You can also see visually when that happens.

I like the cotton thread best for sturdy edges of 12 deg per side plus; Differentiats better, but for very fine edges the rayon thread is a must and even that may be still too thick.

 

[inspecto]

we should put Cliff and Vassili in the same room and lock the door, but first we would have to remove any sharp pointy objects!

 

[nozh2002]

Right now I start testing Buck 110 420HC, Custom Buck 110 BG-42 and Cabela's Alaskan Guide Buck 110 CPM S30V.

I deside to resharpen all of them and after Diamond Products Ultra Fine (1200) which sharpen all of them without any difference in effort to the poing they easy shave my arm from both left and right sides of the blades I measured sharpness.

420HC

Initially tests went wrong: 120, 80, 180, 220, 180, 240, 240, 240, 230, 300. Cleaarly I has wier edge and so I stroke few times both sides again on same diamond stone. After this, results were:

100 ++
110++
120+++
130++++++
140++X+++
150+++++
160++
170+++
180++
190
200+

140 - final sharpnes bit closer to 130 then to 150.

BG-42

100 ++
110 +
120 ++++++
130 ++++++X++
140 ++++++
150 ++
160 ++++
170
180 +

130 - near 140

CPM S30V

90 +
100 +
110 ++
120 +++++
130 ++++++
140 ++X+++
150 +++++
160 +++
170 +
180
190
200 +
...

140 again.

So for me this means that:
1. Method is working again.
2. It is proven that there is no difference in sharpening effort for cheap and expensive steel - does not really matter what steel it is, result will be same if you are sharpening with sharpening equipment. Actually I had more problem with 420HC because of wier edge then with premium CPM S30V or BG-42.

Thanks, Vassili.

To address Cliff's concerns, I increase number of tests to 31.

 

[Cliff Stamp]

I hope that somebody else will try this method also (this is the point of all this) and then we may see more independent results.

As I noted, I have been doing that for over six years now and many other people have adapted similar results. You can find a massive amount of data of this type already performed. I compiled a basic overview of it a few months ago which includes a reference to your work.

To address Cliff's concerns, I increase number of tests to 31.

My concerns have nothing to do with the number of points.

-Cliff

 

[nozh2002]

I do not see reason to do this additional preparation. For me results speaks for itself - last measurement shows pretty good gaussian distribution on different blades. For me it is enoght to think that everything is fine with mesuarement tools.

Similar pattern I noticed with Spyderco Military. If we keeps in mind that edge may blunter not evenly and it is impossible to cut always in same point then we may expect distribution not being perfectly gaussian, but just sharpened edge when condition is more or less even show perfect gaussian distribution (see this on Military initial condition and last examples with just sharpened Bucks).

I consider this as a good sign that equipment does actually working. I tryed with electronic scales, but they was to slow that in this case results was really pointless, and numbers actually behaves crazy as result. In this case as you may see by results somehow all parts fits each other.
-----------
I can imagine that yes that sinusoidal pattern is just a noise, but I also can imagine that it is bending-then-breaking edge nature, again we shall see when more tests will be done. But there is here background for this or that hypothesis.

Thanks, Vassili.

 

[nozh2002]

After 6 microns diamond powder on leather.

Buck 110 420HC

70 +
80 +++++
90 ++++++++
100 +X++
110 ++++++
120 +++++
130 ++

Buck 110 BG-42

70 ++
80 ++++++++
90 +++++X+++
100 +++++
110 +++
120 ++++

Buck 110 CPM S30V

80 +++
90 +++++++
100 ++++X++
110 ++++++
120 +++++++
130
140 ++

Again it is pretty gaussian distribution.

Again BG-42 has better sharpnes on 1 point - is it pattern?

Again I sharpened them with same effort for each and results is same with entry level steel and premium steel. I would say that premium steel shows slightly better disstribution whiah may indicate better consistence and better responds to sharpening.

Thanks, Vassili.