throwing powder statistically

[jmorris]

What causes the variation? Is it a given volume of the powder weighs differently or a different number of kernels gets crunched up each time a charge is thrown?

If its the 2nd one, more throws will just make things worse. Kind of like opening one or two of the leafs in a Lyman 55 measure all the wasy vs just cracking all three a small amount. One maximizes exposure, the other minimizes it.



This I have tested and all 3, reducing the number disrupted is both smoother and more accurate.

Anyway, it should be easy to test weight wise and if you do that, I would be curious if for some reason multiple throws are more consistent, how the increased surface area due to the more fractured kernels helps or hurts consistency upon firing.

 

[taliv]

i'm taking it on faith that a given volume weights the same. i think that's a fairly safe assumption. i also don't believe it's crunching the powder as i only crunch one about one out of ten throws. most of the time it's super slick. i think it's simply variation in volume as powder falling through an aperture into a barrel and it doesn't always fill it up, but i could be wrong

 

[denton]

Several people have posted the opinion that multiple throws will increase variation rather than reduce it.

This proposition requires that the errors in individual throws tend to accumulate in one particular direction. The opposite is true. They will almost always partially cancel instead of completely accumulating.

Shooters are geared to think in terms of "extreme spread", which is what statisticians call "range". And the natural tendency is to think in terms of simply adding ESs. But it doesn't work that way.

The model that works is to think in terms of standard deviations. And the standard deviation of the mean (and by simple extension, the sum) of samples drawn from a population is (standard deviation of population)/(square root of sample size). That's been well understood for over a century, and is the basis of the One Sample T Test, which dates back to about 1920. Thank you William Sealy Gosset.

 

[Laphroaig]

The unit density of granular materials is dependent on the packing arrangement of the grains. The grains assume a denser, more efficient arrangement when subjected to vibrations. Think of a box of Cheerios that is full when it leaves the factory, but only 7/8 by the time it gets to your house. So you dump powder into the barrel and take your chances as to how the grains will arrange themselves. Spherical ball powders apparently are able to get packed more efficiently when dumped than stick powders and therefore can be thrown more consistently. I've heard of loaders who vibrate the measure in various ways to get the kernals to pack in a denser, more consistent arrangement before they throw the powder.

 

[Bill M.]

What if you forget and throw 5 instead of 4. I just do not like the idea. Weigh each charge.

 

[styles]

probability vs distribution

You'd need a pistol measure for a 7.75gr x 4 test.

 

[jmorris]

i'm taking it on faith that a given volume weights the same. i think that's a fairly safe assumption. i also don't believe it's crunching the powder as i only crunch one about one out of ten throws. most of the time it's super slick. i think it's simply variation in volume as powder falling through an aperture into a barrel and it doesn't always fill it up, but i could be wrong

I don't think that is true, if it were the case, volume measures would be all we would use. For solids or liquids, that are consistent in composition I think they are much better though.

Gun "powder" defines a large group of substances though and some meter better in volume measures than others. Some of the ones that are considered the best for some uses meter awfully in volume measures.

I know when I was involved with growing silicon crystals we not only vibrated the granular silicon but also flowed argon from the bottom, up through the column, to make it as consistent as we could.

 

[jmorris]

Several people have posted the opinion that multiple throws will increase variation rather than reduce it.

This proposition requires that the errors in individual throws tend to accumulate in one particular direction. The opposite is true. They will almost always partially cancel instead of completely accumulating.

What if we just used a tiny collator with linear accelerator out feed so we could optically count every kernel, as it exited, so we had the exact same number every time?



Would make for very precise charges, if every kernel weighed the same all the time but they don't. So you wind up with a similar problem that volume throwers have and the reason the "trickling to weight" tools and machines exist but throw 240, .1gn charge machines don't.

After all, if two 12 gn throws were going to be more accurate than one 24, 4 x 6 grains would be even better but 240 tenth of a grain charges would be much better and the 1200 count of .02gn kernels would rule.

 

[Varminterror]

If your error is random, and not from shifts in the process, the answer is that 4 dumps of 10 grains will typically be more precise than one dump of 40 grains. The standard deviation comes down by the square root of the number of samples. The square root of 4 is 2, so your variation will come down by a factor of 2.

Do you believe this to be true considering the form of the powder measure?

I struggle to believe such for powder drops due to the relative proportionality and form packing factors of the system and substrate. Considering the proportionate volume of transitionary regions, for example the ends of the roughly cylindrical drop cavity, vs. a relatively standardized bulk middle section.

Given the same form of powder, I would expect the larger relative proportion of straight section being utilized will yield a considerably more repeatable volume than one which is relatively more dominated by “ends”.

The layman’s example of course being a hamburger bun (standing on edge) versus a loaf of bread. A hamburger bun is effectively all heels, asymmetric ends whereas the longer form loaf is made up predominantly by uniform and symmetrical slices. In this case, “heels” are analogous to end effects, and uniform slices are analogous to relatively uniform packing within the regular cross-section. Proportionately more “heel” = proportionately more opportunity for induced error by inconsistent fill in the transitionary zones of the ends.

Referencing of course, how many engineering school modeling systems are done in “infinitely long cylinders” to dilute to the point of elimination any consequence due to transitionary end effects and non-conformities. I’ve had to break a lot of young engineers of these bad habits when we discuss packed bed reactors as well as scrubbing and distillation columns.

My contention would be that although we may statistically expect to get lucky such a short first throw might be marginally corrected by a long third throw, we MAY be increasing the relative error of our system by eliminating our proportion of consistent cross section - aka, we’re reducing the influence of the part of the system that varies the least in every individual throw stroke, and increasing the influence of the part of the system which induced the most error. Exacerbated by the cutting/crunching effect @jmorris mentions above - reducing net regularity in our fill volume cross section AND introducing proportionately greater variant form substrate.

 

[joneb]

Stick powders can be a pain, the more they are cut into smaller pieces by your powder throw the denser or heavier they get per throw.

 

[denton]

Do you believe this to be true considering the form of the powder measure?

I struggle to believe such for powder drops due to the relative proportionality and form packing factors of the system and substrate. Considering the proportionate volume of transitionary regions, for example the ends of the roughly cylindrical drop cavity, vs. a relatively standardized bulk middle section.

Given the same form of powder, I would expect the larger relative proportion of straight section being utilized will yield a considerably more repeatable volume than one which is relatively more dominated by “ends”.

The layman’s example of course being a hamburger bun (standing on edge) versus a loaf of bread. A hamburger bun is effectively all heels, asymmetric ends whereas the longer form loaf is made up predominantly by uniform and symmetrical slices. In this case, “heels” are analogous to end effects, and uniform slices are analogous to relatively uniform packing within the regular cross-section. Proportionately more “heel” = proportionately more opportunity for induced error by inconsistent fill in the transitionary zones of the ends.

Referencing of course, how many engineering school modeling systems are done in “infinitely long cylinders” to dilute to the point of elimination any consequence due to transitionary end effects and non-conformities. I’ve had to break a lot of young engineers of these bad habits when we discuss packed bed reactors as well as scrubbing and distillation columns.

My contention would be that although we may statistically expect to get lucky such a short first throw might be marginally corrected by a long third throw, we MAY be increasing the relative error of our system by eliminating our proportion of consistent cross section - aka, we’re reducing the influence of the part of the system that varies the least in every individual throw stroke, and increasing the influence of the part of the system which induced the most error. Exacerbated by the cutting/crunching effect @jmorris mentions above - reducing net regularity in our fill volume cross section AND introducing proportionately greater variant form substrate.

An interesting question.

None of my powder measures seem to suffer from such an issue. I did check one of them once and found the random error to be constant across the typical range of charges.

 

[Nature Boy]

This is what us anal retentive types need for powder dispensing