It is not good for my ego to have the predictions mostly correct again, but this time there were things that surprised me a bit.

I have included a sharpening angle of 10° which I never use in praxis because it is not recommended for the N690 steel due to reduced edge retention at that angle (tendency to chipping), but that would not be a problem in this test and it is a data point of knowledge in case I make some carbon-steel sushi knives or razors in the future.

Now the boxplot:

So I got the prediction about how the cutting force will rise mostly correct. Mostly, not completely.

ANOVA test has found no significant difference between the first four angles but I am sure there would be one if I had performed more measurements and/or refined the testing method. The Lookandsee test does indicate a slow rise in cutting force from around 25 gf to around 50 gf.

The jump at 30° is a bit more sudden than I expected. I suspect that it is a fluke. And then the rise at 40-45° was a lot less than I expected.  It seems that the 90° cutting edge is still significantly better than no cutting edge, which would be somewhere around 3-4 times worse with a cutting force of around 1000 gf. I did not expect that. The best-fit function is quadratic. This is less drastic than the predicted exponential growth, although still significantly faster than simple linear growth.

So in conclusion, it does appear that my opinion that whilst there is a difference at angles 10-25°, it is not big enough to matter for casual knife users is substantiated. The angle 30° performed slightly worse than I expected, and the angle 45° performed significantly better than I expected.

I am going to think about all this some more and then I decide how to proceed from now on.

Notwithstanding dangerousbeans’s comment at the last article, I did find in my steel offcuts pile a blade that broke after it was nearly completely finished, with etched logo and all. That means learning about the sharpening properties of exactly the steel that I use in exactly the state it is in a finished product. In different steel, the results might come out a bit differently, which means I am mightily glad that I could do my tests on this – an absolutely ideal testing specimen.

I have sharpened the blade stub at a 15° angle which is the angle at which I usually sharpen kitchen knives, and always those that I make from N690 steel. (When tasked with sharpening store-bought knives from unknown steel I occasionally sharpen them at 20°, especially if it is clear from the state of the knives that the customer is not particularly careful about their use.)

I established the bevel with 120 grit and then I progressed from 180 and 240 grit Zirkorund and then Trizact belts (in the evaluation translated into grit equivalents) from A 65 all the way to Trizact A6. I only differed from my usual sharpening procedure in one way – I used fresh belts, instead of old ones. Normally for sharpening, I am using old belts because sharpening is extremely rough on the belt and destroys it very quickly.

The measuring method is wildly imprecise – the testing thread is not homogenous, the angle at which I put the blade on it is not always precise, I do not always hit the center, I am not pushing with constant speed, the kitchen scales do not renew the measured value with sufficient frequency and probably several other variables. I have experience with such measurements from my previous job though and there are ways to get relatively reliable, reproducible, and usable results even so. One of those ways is to make lots of measurements – that is why I took 12 measurements, discarding two of the most egregious outliers and making the evaluation with the remaining 10. There are mathematical methods for discarding outliers but for my personal purposes, the Lookandsee method suffices. (Thirty or fifty measurements would be better, but I am not going for exact values for individual grits, I am going for a comparative assessment between those grits. Anyway, not going to write a boring lesson about measuring).

Here is the boxplot of those ten measurements per each grit:

And it looks like my prediction from yesterday was bang-on. Which is satisfying to my ego, but also boring in a way. It would be much more exciting if it were different.

What you see here is the cutting force going rapidly down from ~270 g at 180 grit to ~ 75 at 400 grit. At this point, the blade is capable of cutting freely hanging paper. Then it falls some more to ~40g  at 800 grit and more or less stays there till the end (the knife is shaving-sharp at these stages). The slight increase at 1200 is a fluke that would most probably go away if I have made more measurements and/or invested time and resources in refining the method. I made ANOVA test and there is no statistically significant difference between the last three fine grits.

So in conclusion, sharpening knives beyond 1000 grit indeed appears to have very little practical value. At 800 grit the blade is already shaving-sharp and polishing the edge further only costs more time without noticeably improving its cutting capabilities. With more precise measuring method there might be a difference, but it would be very, very tiny. I think that I can replace the last two belts with a leather belt infused with stropping compound and get the same result. And since time is money, I will do exactly that.

I hope to use my new sharpmeter to get some knowledge about, well, sharpness. And since I am going to be playing at science whilst doing so, I have decided that I will write down the predictions for my tests. The tests will not be blind, because I will be doing both the sharpening and testing and there will still be some subjectivity to these tests, but nothing is perfect. I am doing these tests to gain some knowledge and I will share that knowledge for free but there will inevitably be bias.

The first test that I intend to perform is the influence of grit on edge sharpness. I think that after establishing the bevel with 180 grit the cutting force will go down significantly with the next steps, but it stops changing significantly above 400 grit. My reasoning for this is the fact that it is possible to get a knife shaving-sharp with ca 320 grit stone and at 320-400 grit usually the wire edge/burr falls off. I think that I have mentioned in the past in comments somewhere – either in Marcus’s place or here – that going above 1000 grit in sharpening makes little sense function-wise, although I cannot find the comment now. I will go up all the way to trizact A6 belt (the equivalent of 2000-2500 grit) in the experiment.

The second test that I intend to take is the influence of the sharpening angle. There was a heated debate between me and Marcus on this issue a while back -click-  and I really want to test it (caveat from the first paragraph applies doubly). I expect the force needed to cut the thread to rise exponentially, i.e. slowly from 10-25°, then some more for 30° and even more for 40° and again even more for 45°. I won’t test sharpening angles steeper than 45° because it makes no sense IMO since a 45° sharpening angle means a 90° edge. I know from praxis that knives sharpened at 15°, 20°, and 25° can be shaving-sharp. I do not know much about the 30° angle, since that is extreme and I only sharpen hatchets and axes at that angle and I never even tried to get those to shaving-sharp. They do cut paper though.

So, sometime this week I shall heat the workshop again, sharpen some steel offcuts (probably pieces of old hacksaws) and go measuring.

It is freezing here and I still do not have the slightest inclination to do something useful. But I need to heat the workshop occasionally to prevent it from completely freezing. Not that it would be super bad, except that maybe ice forming in the cooling receptacle near the grinder could damage it. Anyhoo, yesterday was a workshop heating day. I could not do anything super useful – it took me hours to raise the temperature to a bearable level and soon after I stopped feeding the stove, the temperature got down quickly again. But I could use the time to do something small and simple so I have made a device for measuring knife sharpness (I did not invent the concept, I saw the principle somewhere on the interwebs sometime ago).

It is simple and consists of two main parts. One part is a board with four legs and a 35 mm hole in the middle. A tiny table that can be put over my kitchen scales.

The second part is a small wooden cylinder with a cutout and two screws on each side of it.

I can span a thin thread between the two screws and when the cylinder is put on the scale through the hole in the middle of the tiny table, I can cut the thread safely for me, the scale, and the furniture.

The main downside is that my kitchen scale does not have a “Hold” or “Max” function so the measured values are not super precise. Another problem is the used thread – a very thin fishing line would be probably better since this one has a tendency to get damaged during spanning. Or perhaps a very thin copper wire – I might try to extract some strands from leftovers from speaker or ethernet cables. But when being very careful with spanning the thread and doing the cut slowly and carefully, the setup gives useable results and I did learn some things.

Here is a boxplot of 10 measurements with the three cutting implements in the photos – a fresh razor blade, a paper-cutting-but-not-shaving-sharp sharp knife, and a blunt table knife (I used the non-serrated part which is about 1 mm thick),

The razor had an average cutting force of ~6 g, the sharp-ish knife ~60 g, and the blunt knife ~ 1000 g. The less sharp the blade the bigger the spread but even with sometimes ridiculous outliers, there is a definitive and statistically significant difference between these three and it does give me some information and opens future opportunities. I would like for example to examine the relationship between sharpness and sharpening angle, to get some hard data to back up my opinion that anything between 10-30° works just fine. My prediction is that the relationship is not linear and as the angle gets steeper, the cutting ability gets exponentially worse.

I did learn one thing – my “Not a Masterpiece” was sharpened at a 20° angle and the only knife that actually scares me – the bigger knife from the two-knife applewood set – was sharpened at 15°, yet both measured within the same range as the razor. Although the applewood knife completely failed to register on the scale one time, thus my suspicion that it is the sharpest knife I made so far might be true.

Of course this only tests edge geometry, not blade geometry. I could use a similar setup to test the influence of blade geometry on cutting force too, but I do not have a reliable medium to do so yet. All things that I have thought of so far are either expensive (cork, rubber, silicone) or have highly inconsistent properties (fruit & veggies). But maybe I will think of something to test blade gomtry too.