Yrrah Method The Machine and User Experiences ~~ Part 5

Kind of what I expected the outcome to be like although the cell #7 was a bit better than I expected. Keep up the good werq for all of us lazy dudes and gals out here.

Click to expand...

#7 is going to be the preferred head size for that barrel I expect. This explains that flier in every group. You get two that have the same head size and skirt diameter and then you get one that is half a thousandth large or small and that's your flier. You had two in the same hole and then that one opened the group to half an inch, which was still good but ...

Harry and I are chatting. He sorts his pellets on a board that allows a 180 degree roll and he sorts them into groups the length of the pellet. He uses a feather to start the pellet rolling. His implementation is more precise than anything I have ever seen but it would be hard for someone to implement HIS method in a machine. He can document.

Here is a picture that Harry gave me with permission to share ... This shows Harry's setup and is his implementation the YRRAH method.. This is how accurate a roller can be:

This proof has been checked by Bob Sterne. I think it is a good proof. If someone else can correct it I will thank them. "If the machine is working correctly AND the pellet is rolling correctly, two pellets with different dimensions can not roll on exactly the same radius. Even pellets having the same head to skirt ratios but of different lengths OR FROM different sections of the cone, even overlapping sections, will not roll to the same place." I have already had to fix typos in this, if you see one point it out please.



This calculation also shows that this test is very sensitive to any change in the ratio of head to skirt or length for that matter but I have not calculated that so can not quantify it. We see that a 0.01mm change in skirt diameter makes approximately a 500mm difference in the rolling radius. That seems enormous to me and I'd think my math is wrong. By all means check it and correct me. I am old and my mind does not work like it once did. Barring that correction, this makes it clear that pellets which roll to the same spot are very similar. If the problem were linear ( I don't think it is but have not wandered down that track )... but if it were a linear error we are talking about being able to detect errors in skirt to head diameter that are not measurable by any other means that is remotely as cost effective.. ... which dove tails nicely with the error rates I am seeing now on the machine.

Pellets which skid or will not roll (damaged pellets) will generally fall much further down the table than pellets which roll perfectly. Pellets should roll as slowly as practicable. The slope I am currently using is 1 mm drop for 25 mm in run (one mm per inch). If you are building a machine I'd encourage you to try one in thirty. I believe you'll get better results but you'll get a lot of "no rolls".

A mechanical release mechanism of some sort is imperative if you want to maximize your result. You can roll by hand if you are willing to roll each pellet multiple times and take only those which hit the same spot multiple times in a sequence. So for example, if you were rolling by hand and your error rate was 12% by hand, then if you rolled the same pellet 4 times into the same spot the effective error rate would be 12% then 12/2=6%; 6/2=3% and finally 3/2=1.5%. To be honest if I were releasing by hand I would not batch pellets which had not rolled 4 times into the same slot in 4 consecutive tires.

If they are not a keeper, no problem. Put them back into the tin. As you remove similar pellets from the tin, those which remain in the tin will become "more similar" than they were when you started. So you need fifty pellets, you roll up fifty and put the ones that did not sort back in the tin. Next time you roll fifty, your ZERO will change but your group size should still improve. That said, there will be an optimal head size your rifle prefers. Those pellets are the ones you want to identify when you start rolling out a tin so that you can set those aside for when you go to EBR ... or that FTT match ... or when you want to head shoot that chipmunk at 100 yards, rather than gut shoot him.

Some will say I have wasted your time. I hope not. I intend to keep posting my results here at the bottom of this thread, whether or not they support my current opinion.. I got you into this, if I got it wrong I will get you out.

Thanks guys for your patience.

I think I am about done. I'll make a little video that gives you the dimensions of my roller and link it here when I work up the desire to spend another hour on this . Right now, my grass needs cutting and I've about extracted all the fun I can on this project.

EDIT: 

• Please note the numbers I used in the graphic were selected for the proof, to simplify the math, not to be realistic. 
• Most of the pellets I have looked at will roll a considerably smaller radius than the numbers mentioned in the graphic. That is because the difference in head to skirt diameter is RARELY less than a couple of thousandths OF AN INCH, say 0.051mm to 0.102 mm literally an order of magnitude bigger than I used in the graphic. That would reduce the roll radius to several inches vice a meter. 
• I just realized the change in skirt size I introduced on the graphic represents a 0.02mm DIAMETER change so something like two thousandths of an inch. That produced a delta in the roll radius of about one third (1000mm vice 1500mm).
• We can extrapolate from that approximately what resolution we are actually achieving with a roll table. If we are rolling into a cell size of 10mm with the notional pellet we are discriminating skirt diameters which vary by about 0.001 mm or 0.00005 inches by my math. That is probably going to show up a problem with alignment between head and skirt as well... and if you get the roll speed and tilt just right pellets that are only out of balance (pellets which have the correct outside dimensions but which have cavities that are off center) will likely fall long as well due to slippage on the glass. 
• The difference in roll radius for two pellets 6mm tall having r1`= 2.5mm and r2 = 2.545 mm and 2.546 mm respectively is almost exactly 7.25 mm. With a base roll radius of 333 mm ... We are discriminating by 1/1000 mm. That is 0.000039 inch (about four one hundred thousandths) with a PERFECT roll.

Here is an exception to the proof.

More discussion with R. Sterne over at AGG.

https://airgunguild.com/ask-bob/geometric-proof/ over

An image I received from @yrrah showing data he collected. The SD of that data is less than 1 percent of the diameter.

Data Collected Today

Rifle D430L .177, Hawke 4x16

25 yards indoors.

This test basically discovers the limits of my ability and my Diana 430L. I might be able to do better with the 430 given a different design for the aim point. I believe that is indicated by test #3 where I had to use the bottom of the bull as the aim point because I ran out of targets with the diamond. I believe that even though I was tired having already shot 120 pellets, the aim point being different extracted more from the rifle than I was getting by trying to hold that cross hair on that diamond. I will be redesigning that target again. Also that invalidates the data for target #3.

I sorted a tin of H&N FTTs into 9mm cells on the machine earlier this week. When I sort any pellet which rolls off the machine and does not hit a catch box goes into the "no roll" group. Any pellet which falls in the top third or bottom third of the box winds up in the "no rolls" batch. I got about 40 or 50 no rolls in that tin. I sized those to 4.52mm and set them aside.

Pellets that rolled into the three central catch boxes (9mm wide) were separated into groups #1, #2, and #3.

A control group of about 40 pellets was selected from the tin before any pellets were rolled.

I shot the groups in the order presented. There were 150 shots total in the test today. The order was selected to favor the groups I expected to perform worst so that fatigue would not skew the tests in a favorable direction.

The targets and aggregates are below. The data files are attached at the end of this post.

NO ROLL GROUP, Sized to 4.52mm





Control Group





Test Group #1





Test Group #2





Test Group #3





https://www.youtube.com/watch?v=DVGoby1fggY

View attachment 430_hn_1.1628721449.csv

View attachment 430_hn_2.1628721449.csv

View attachment 430_hn_3.1628721449.csv

View attachment 430_hn_ctrl.1628721449.csv

View attachment no-roll-4,52-sized.1628721449.csv